stacks.cdc.gov€¦ · Web vieweAppendix. I. eMethods. A. Study Group Construction and Deductible Imputation Algorithm. To determine employer deductible levels, we used a benefits

eAppendix

I. eMethods

A. Study Group Construction and Deductible Imputation Algorithm

To determine employer deductible levels, we used a benefits type variable that we had for most smaller employers (with approximately 100 or fewer employees). For larger employers, we took advantage of the fact that health insurance claims data are the most accurate source for assessing out-of-pocket obligations among patients who utilize health services. Our claims data contained an in-network/out-of-network deductible payment field. For patients who use expensive or frequent services, the sum of their yearly deductible payments add up to clearly identifiable exact amounts such as $500.00, $1000.00, $2000.00, etc. When even several members have these same amounts, it provides strong evidence that the employer offered such an annual deductible level. It is also possible to detect employers that offer choices of deductible levels when multiple employees have deductibles at two or more levels, such as 20 employees with an annual amount of $1000.00 and 12 employees with $500.00. For employers with at least 10 workers, we therefore summed each employee's in-network deductible payments and number of claims over the enrollment year and assessed other key characteristics such as percentage with Health Savings Accounts. On a randomly selected half of the employer data set that contained our calculated employer characteristics (such as the percentage of patients with deductible levels between $1000-$2500) as well as actual deductible amounts, we used a logistic model that predicted the 3-level outcome of deductible <=$500/$500-$999/>$1000 based on multiple aggregate employer characteristics such as the first and second most common whole number deductible value, the percentage with Health Savings Accounts or Health Reimbursement Arrangements, the median deductible payment, the percentage of employees using services, the employer size, the percentage of employees with summed annual deductible amounts (from claims data) between $100 to ≤$500/ >$500 to <$1000/ ≥$1000 to ≤$2500/ >$2500, etc. This predictive model output the probability that employers had deductibles in the three categories (summing to 1) and we assigned the employer to the level that had the highest probability. If we detected employers that had 10 or more employees with whole number deductible levels both above and below $500 (e.g. $250.00 and $1500.00), we assigned the employers' category as "choice." If 100% of employees had Health Savings Accounts, we also overwrote any previous assignment to classify the employer as a high-deductible employer. We tested the predictive model on the other half of the sample for which we had actual deductible levels (eTable 1). At employers with 75-100 enrollees, we found sensitivity and a specificity of over 96%. The sensitivity and specificity would be expected to be even higher at employers with more than 100 enrollees (because more claims data would be available to provide evidence of deductible levels), but we were unable to test this because the dataset for which we had actual deductibles included employers with generally 100 or fewer enrollees.

Rationale for low- and high-deductible cutoff values: when health savings account-eligible HDHPs came to market in 2006, the Internal Revenue Service set the minimum deductible level for qualifying HDHPs at $1050 (which could be adjusted upward for inflation annually). The range of this minimum deductible during our study period was $1050-$1200. For these reasons, we defined HDHPs as annual individual deductibles of at least $1000 (otherwise health savings account plans would be excluded). In addition, choosing this cutoff (as opposed to e.g. $2000) also improves the sensitivity and specificity of the imputation because this is common deductible level and more enrollees per employer meet this threshold. This cutoff is also a “real-world” deductible minimum that allows the most generalizable results. We did not create a separate imputation algorithm for deductible levels of e.g. >=$2000 due to concerns that a less sensitive and specific algorithm would lead to biased effect estimates and a smaller HDHP sample size. It is important to note that $1000 was the minimum annual deductible level and not the mean deductible level. We cannot calculate the mean deductible level of the HDHP group directly but would expect it to be in the range of approximately $1500 to $2000.

We defined traditional plans as having deductible levels of ≤$500 after determining that a threshold of ≤$250 would lead to an inadequate sample size for the control group. Again, the mean deductible level of the control group members would be lower than $500.

After assigning deductible levels at the employer plan year level, we began with 1,830,665 employer plan years. We excluded 201,230 plan years (11%) that included deductible levels other than only low or only high. Among the remaining 1,629,435 plan years, we excluded 191,519 (12%) that did not have 2 years of continuous enrollment. Finally, from the remaining 1,437,916 employer plan years, we excluded 549,638 (38%) that were not transitions of low deductible to low deductible or low deductible to high deductible. Most of these exclusions were due to employers having high deductibles at their initial appearance in our dataset and remaining in high deductible plans.

Our HDHP group therefore comprised the enrollment years of employers that had a year-on-year transition from low- to high-deductible coverage (from $500 or less to $1000 or more). Some employers had multiple eligible index dates (e.g., multiple low-to-low deductible years or both low-to-low and low-to-high deductible years). In these cases, we randomly assigned employers to the HDHP or control pool then randomly selected one of their index dates (and their corresponding before-after enrollment years).

We identified patients with diabetes age 12 to 64 as defined by detection of 1 inpatient or 2 outpatient diagnosis codes for diabetes (eTable 2), or the dispensing of insulin or at least one oral hypoglycemic medication other than metformin alone, between 6 months before to 6 months after the beginning of members' baseline period.

B. Propensity Score Matching Approach

Propensity score matching assists in generating a control group with a similar likelihood of being exposed to a given “intervention” (in this case, shifting to HDHP coverage) based on measured characteristics when individuals have not been randomly allocated into study groups.1-4 To perform propensity score matching, we first included all employers with at least two years of enrollment. After classifying every employer's index date as before 2008 versus 2008 and later, we used an employer-level model to generate annual employer propensity score quartiles. This logistic model predicted the likelihood of an employer joining a HDHP in a given calendar period (before 2008 versus 2008 and after) based on calendar index date, employer size (<50, 50-99, 100-249, 250-499, 500+); percentage of women; members in income strata, education strata, age strata, race strata, and region strata; employer baseline cost level and trend; average employer Adjusted Clinical Groups (ACG) score; and outpatient copay. We assigned each employer in a given calendar period to a propensity score quartile.

We then identified for inclusion all members age 12 to 64 with diabetes. We classified into these calendar period-quartiles all HDHP group diabetes patients with at least one year of continuous enrollment before and two years after the index date. We used member-level propensity score matching on age, gender, race/ethnicity, neighborhood poverty and education, US region, Adjusted Clinical Groups score, and calendar index month. Based on evidence that propensity score matching on the functional form of the baseline outcome trend closely approximates randomized controlled trial results,5 we also matched on baseline quarterly rates of high-priority primary care and specialist visits, emergency department visits, and hospitalizations, as well as out-of-pocket expenditures. To create the final cohort, we performed this match, within the calendar period employer quartiles, on members who transitioned to a HDHP between January 2004 to January 2011 (allowing a 2 year follow-up ending in December 2012 at the latest) to contemporaneous control members using 1:1 caliper matching without replacement. We used a caliper width equal to 0.2 of the logit of the pooled standard deviation of the propensity score, which has been found to eliminate the majority of bias due to measured confounding variables.4,6

C. Outcomes Measures

High- and low-priority outpatient visits

We applied a taxonomy developed by Fenton and colleagues33 to characterize each office visit among diabetes patients as "high-priority" or "low-priority" based on the primary diagnosis, with high-priority conditions being considered those more likely to benefit from medical care than low-priority diagnoses.33 The classification system was originally derived from the Oregon Prioritized Health Services List, which included an evidence-based ranking of over 700 diagnosis and treatment combinations according to the expected morbidity or

mortality benefits of medical care.33,34 In our baseline sample, the five most common low-priority diagnoses were acute upper respiratory infections, acute sinusitis, acute pharyngitis, allergic rhinitis, and cough, while the most common high-priority diagnoses were diabetes, hypertension, hyperlipidemia, coronary heart disease, and acute bronchitis.

Outpatient disease monitoring measures

Outpatient disease monitoring measures were captured based on Healthcare Effectiveness Data and Information Set specifications, including ≥1 or more annual: primary care visit; hemoglobin A1c, low-density lipoprotein cholesterol, and microalbumin test; and retinal eye examination. Codes used to capture these tests are listed in eTable 3.

Creation and validation of a measure of diabetes acute complications

Overview: We sought to create a measure of outpatient and emergency department visits indicating that a patient had experienced a diabetes complication that can arise when patients defer or skip necessary care. We defined these acute diabetes complications as symptoms or conditions that may be associated with inappropriately deferring recommended or urgently needed diabetes-related outpatient or emergency department care for up to 4 months and that require timely care by medical professionals. eFigure 1 displays the 5-step decision algorithm we used to operationalize this definition on claims-data-based diagnoses. Our general goal was to choose specific primary diagnoses with a high degree of face validity for being possibly related to deferral of appropriate care. Other important considerations were that complications by definition must require urgent or emergent care by a medical professional (i.e., they could not be treated at home) and that certain diagnoses have different meanings when coded in the outpatient or emergency department setting (e.g. “congestive heart failure” is more likely to represent an acute complication if coded as the primary diagnosis in the emergency department versus outpatient setting).

Candidate ICD-9 codes: We developed an initial set of candidate ICD-9 codes using a published list of acute and long-term diabetes complications used to examine the economic impacts of diabetes.7 We also included diagnoses from the Agency for Healthcare Research and Quality’s claims-based Prevention Quality Indicators (http://www.qualityindicators.ahrq.gov/Modules/pqi_resources.aspx) diabetes complications measures which include a smaller number of acute and long-term diabetes complications. The two clinicians on our team (E.M.E and J.F.W.) added three diagnoses with a high degree of face validity for being potentially related to gaps in appropriate diabetes and diabetes-related care: hypoglycemia, influenza, and pneumonia.

Classification for inclusion in final measure: The two clinicians on our team (an endocrinologist [E.M.E.] and a general internist/ urgent care physician [J.F.W.]) then independently applied the 5-step decision algorithm shown in eFigure 1 to each of the ICD-9 codes, separately classifying them as a potential acute complication when coded as the primary diagnosis in either the outpatient or emergency department setting. We excluded any diagnoses (the “No” branches as shown in eFigure 1) for which the answer to any of the 5 steps was uncertain or ambiguous, in order to prioritize specificity over a sensitivity. After each clinician classified the diagnoses independently, we selected for inclusion in the final measure only those ICD-9 codes where both clinicians agreed that the diagnosis represented a recent acute complication. We therefore ended up with a list of 64 outpatient and 89 emergency department acute diabetes complication ICD-9 codes (eTable 4).

Creation of acute diabetes complication visit measure: We then captured the primary diagnosis coded by clinicians (using evaluation and management CPT codes and the associated first ICD-9 codes) at all outpatient or emergency department visits. We separately flagged outpatient and emergency department visits that had a primary diagnosis on the respective list shown in eTable 4. Because we did not want to count repeat visits for a single complication episode, we required an interval of 10 days after a first detected episode of a complication. For example, if a patient was diagnosed with cellulitis on day 1 then was seen for this again on days 3 and 7, we would only count this as one episode rather than three.

Validation of acute diabetes complication visit measure: We validated two aspects of our acute diabetes complication measure: (1) the association of gaps in outpatient care (including prescription drug use) and emergency department care with subsequent acute complications and (2) the association of diabetes

complication visits with subsequent hospitalizations. Our basic respective hypotheses were: (1) Because diabetes patients who are not diet controlled should have essentially all enrollment days “covered” by medication on hand, and/or should present with concerning symptoms within days (e.g. within 7 or fewer days), any gaps in care of 7 or more days would be associated with a higher risk of subsequent acute complication visits than periods when there were no gaps in care; and (2) Because our definition of complications comprised symptoms and conditions that require urgent or emergent medical attention, acute complications would be associated with an increased risk of subsequent hospitalizations.

Association of gaps in care with complications. We first identified gaps in outpatient and emergency department care as any day when a diabetes patient, who had received at least a 1 month supply of a medication related to preventing or treating diabetes or its complications, subsequently had days without such medications or without an outpatient or emergency department visit. The classes of drugs we included as potentially related to preventing or treating diabetes or its complications, based on AHFS categories, were: antibacterial (AHFS: 0812), antifungal (0814), autonomic (12), blood/coagulation (16, 20), cardiac (2404), lipid lowering (2406), antihypertensive (2408), vasodilator (2412), analgesics/antipyretics (2808), antidepressant (281604), antipsychotic (281608), ear/nose/throat (52), insulin (682008), oral antidiabetic (6820), and skin agents (84). Among patients who had at least one gap between 7 to 120 days long (“gap group,” n=297,413 of 364,020), we then randomly selected a sample equal to half (n=182,010) of our overall pool of diabetes patients. We randomly selected one gap (i.e., range of dates with no diabetes-related care) to analyze for each of the randomly selected 182,010 patients. We propensity score matched (caliper; 1:1 ratio) each member in the “gap group” to the other half of the pool of patients based on date of diabetes diagnosis, date of first diabetes-related prescription, date of end of enrollment, age group, AHFS morbidity score, race/ethnicity, education, gender, poverty, and US region of residence. This left us with 158,916 “gap group” patients and their matched controls. Because the matched control patient would not necessarily have a “non-gap” period during the exact contemporaneous days that the “gap group” had gaps, we restricted the sample to the pairs where the patient with a gap had a matched control with an exact contemporaneous non-gap (n= 39,828 in both groups). We used logistic regression to estimate the odds of an acute complication on the day following a gap in care compared to the day following the contemporaneous period of the matched subject. eTable 5 shows the raw frequencies of acute complications following gaps and non-gaps, corresponding to an odds ratio of 7.5 (4.9, 11.5). (We also performed the same analysis, but regardless of whether the matched “potential non-gap” member had a contemporaneous “non-gap” or not, but ensuring enrollment during the period of interest, finding an odds ratio of 10.1 [7.3,14.1]).

Association of acute complications with subsequent hospitalizations. To confirm that the acute diabetes complications visits we selected were indeed high acuity, we used logistic regression to estimate the odds of hospitalization within 3 days after the complication visit compared all other visits. We performed this analysis separately for outpatient visit acute complications (compared with non- acute complication outpatient visits) and emergency department acute complications (compared with non- acute complication emergency department visits). eTable 6 shows raw frequencies, and the corresponding odds ratios were 4.10 (3.98, 4.23) and 3.02 (2.96, 3.08), respectively.

D. Covariates

To estimate comorbidity, we applied the Adjusted Clinical Group (ACG) algorithm to members’ baseline period. The algorithm uses age, gender, and ICD-9-CM codes to calculate a morbidity score and the average of the reference population is 1.0.8 Researchers have validated the index against premature mortality.9

To derive proxy demographic measures, the data vendor linked members’ most recent residential street addresses to their 2000 US Census block group.10 Census-based measures of socioeconomic status have been validated11,12 and used in multiple studies to examine the impact of policy changes on disadvantaged populations.13-15

We classified members as from predominantly white, black, or Hispanic neighborhoods if they lived in a census block group (geocoding) with at least 75% of members of the respective race/ethnicity. We then applied a superseding ethnicity assignment if members had an Asian or Hispanic surname,16 and classified remaining members as from mixed race/ethnicity neighborhoods. This validated approach of combining surname

analysis and census data has positive and negative predictive values of approximately 80 and 90 percent, respectively.17

E. Analysis

For the high priority primary care and specialist visit outcomes, we first fit interrupted-time-series models18 in order to both visually display monthly trends and confirm that the study groups did not have differential baseline trends, a key assumption of difference-in-differences analysis. We used difference-in-differences analysis to examine changes in annual high-priority outpatient visits and disease monitoring measures. We applied generalized estimating equations19,20 models with a negative binomial distribution for outpatient visits and a binary distribution for disease monitoring measures, controlling for (as appropriate) age, gender, race/ethnicity, education level, poverty level, US region, ACG score, employer size, and index date. The term that estimated the HDHP effect was the interaction between study group (HDHP versus control) and study period (follow-up year versus baseline). We applied marginal effects methods21 to estimate adjusted rates of measures and to estimate adjusted absolute and relative baseline-to-follow-up changes in the HDHP group versus the control group.

To examine study group differences in time to first outpatient and emergency department visit for acute complications, we used separate Cox proportional hazards regression models for the baseline and follow-up periods, adjusting for the same covariates as above. The term of interest from the regression models was a binary indicator of membership in the HDHP group. This term generated adjusted hazard ratios (aHR) of the HDHP group compared to the control group in the baseline or follow-up periods.

To analyze annual changes in complication visits and associated subsequent 7-day total costs, we applied aggregate-level segmented regression to adjusted cumulative rates. This involved first using generalized estimating equations19,20 with a Poisson distribution to model monthly visit rates and costs in both study groups, adjusting for the covariates above and accounting for clustering at the person level.22 We then applied marginal effects methods21 to calculate monthly rates in both groups that were fully adjusted for the preceding covariates. We generated cumulative monthly rates23,24 from these adjusted monthly rates, and plotted the cumulative control and HDHP group rates before and after the index date. This approach allows visualization of changes in rare outcomes that gradually accrue over time and prediction of cumulative rates at a given follow-up time point based on the baseline trend in the monthly cumulative points. We modelled cumulative HDHP and control group trends using aggregate-level segmented regression,18 adjusting standard errors for autocorrelation. The regression models included intercept, baseline trend, trend change, and quadratic trend change terms for the HDHP and control groups, and were included in final models using backwards elimination with a threshold of p<0.05. Using marginal effects methods,21 we estimated absolute and relative changes in the HDHP group compared with the control group at the end of follow-up versus the end of baseline using the above segmented regression terms.

Using the same methods and outcomes, we conducted subgroup analyses stratified by low- and high-income and morbidity and examined HSA members and their matched pairs. We also assessed several other subgroups of interest including by other income cutoffs (residing in neighborhoods with under 10%, over 5%, and over 20% of households below the Federal poverty level), and residents of predominantly white and non-white race neighborhoods.

II. eResults

A. Changes in Out-of-Pocket Exposure

The overall, low-morbidity, high-morbidity, high-income, low-income, and HSA diabetes HDHP groups experienced increases in mean out-of-pocket medical expenditures of 49.4% (40.3%,58.4%; absolute: $374.6), 56.8% (45.8%,67.8%; absolute: $292.0), 40.9% (31.5%,50.4%; absolute: $448.8), 48.4% (37.2%,59.6%; absolute: $361.8), 51.7% (38.6%,64.7%; absolute: $400.4), and 67.8% (47.9%,87.8%; absolute: $463.0), respectively, relative to controls in the year after transitioning to HDHPs (eTable 7). Out-of-pocket obligations for hemoglobin A1c, LDL-C, and microalbumin tests increased from a mean of $1.2-$1.4 at baseline among HDHP members to $2.2-$4.8 at follow-up (eTable 8). Primary care visit costs increased from $15.4 to $23.3-$26.8 from baseline to follow-up among HDHP members, while specialist visits averaged $23.3 at baseline and approximately $42 at follow-up.

B. Utilization and Disease Monitoring Measures

Interrupted time series analyses demonstrated no statistically significant baseline trend differences between the HDHP and control groups in high priority outpatient visits for all subgroups (implying validity of difference-in-differences estimates; eTable 9 and eFigure 2) except primary care visits among low-morbidity members.

C. Health Outcome Measures

The overall HDHP diabetes cohort experienced a follow-up period delay in the time to first outpatient complication visit compared with controls (adjusted hazard ratio [aHR], 0.94; 95% CI, 0.88-0.99; eTable 10 and eFigure 3) that was not present at baseline (aHR, 1.01; 95% CI, 0.93 to 1.09). eTable 10 also lists all other aHRs from subgroups of interest. Total annual ED complication visits and complication episode expenditures increased by 8.0% (95% CI, 4.6%-16 11.4%) and 5.6% (95% CI, 3.8%-7.3%), respectively, in the overall HDHP group compared with the controls (eTable 11 and eFigure 3). Corresponding changes in these ED outcomes among HSA HDHP members were 15.5% (95% CI, 10.5% to 20.6%) and 29.6% (95% CI, 19.0% to 40.1%; eFigure 4 and eTable 11). eTable 11 also lists all other total annual ED complication visit and complication episode expenditure findings from subgroups of interest.

D. Other Subgroups and Sensitivity Analyses

In analyses of other subgroups of interest, HDHP impacts on high-priority visits (eTable 12) and disease monitoring measures (eTable 13) among key HDHP subgroups did not differ substantially from the overall cohort. Alternate cutoffs for the income group definitions were generally consistent with the hypothesis that HDHP groups with lower income experienced greater adverse outcomes (eTables 10 and 11; eFigure 5), and non-white neighborhood residents also experienced substantial increases in acute complication measures. Restricting to members age 18-64 yielded similar outpatient visit, disease monitoring, and time-to-visit results compared with the age 12-64 cohort, but increases on acute complication measures were generally more pronounced (eTable 14-17).

Analyses of the top 5 most common emergency department complication visits demonstrated effect estimates with similar directions as in the main analysis except among HSA-eligible HDHP members (the smallest subgroup), among whom we detected no statistically significant changes (eTable 18).

Page 6

eTable 1. Validation of deductible imputation algorithm.

Gold Standarda=high-deductible (number of members)

Gold Standard=low-deductible (number of members)

We imputed high-deductible 611,541 14,335We imputed low-deductible 24,017 465,120

High-deductible Low-deductibleSensitivity 96.2% 97.0%Specificity 97.0% 96.2%Positive Predictive Value 97.7% 95.1%aGold standard was a benefits variable specific to each employer and obtained from the health insurer via the data vendor.

eTable 2. Diagnostic and medication codes and used to define the diabetes cohortCode Description ICD-9-CM, DRG, or AHFS CodeTo include in denominator:

Diabetes diagnosis 250.0-250.93Polyneuropathy in diabetes 357.2Diabetic retinopathy 362.0Diabetic cataract 366.41Diabetes mellitus complicating pregnancy 648.03, 648.04Uncomplicated diabetes, age over 351 294Uncomplicated diabetes, age 35 and under1 295Diabetes with MCC2 637Diabetes with CC2 638Diabetes without CC/MCC2 639

To exclude from denominator:Polycystic ovary syndrome 256.4Other specified disorders of pancreatic internal secretion 251.8Poisoning by adrenal cortical steroid 962.0

1Used before 10/1/2007; 2Used on or after 10/1/2007 Abbreviations: ICD-9-CM, International Classification of Diseases, 9th Revision, Clinical Modification; DRG, Diagnosis Related Group; AHFS, American Hospital Formulary Service

Page 7

eTable 3. List of procedure and diagnostic codes used to create disease monitoring measures Code Description CPT, HCPCS, LOINC, or ICD-9-CM Code

Hemoglobin A1c test 17855-8, 17856-6, 3044F, 3045F, 3046F, 3047F, 4548-4, 4549-2, 59261-8, 8046F, 8047F, 83036, 83037

LDL-C test 24331-1, 12773-8, 13457-7, 18261-8, 18262-6, 2089-1, 22748-8, 3048F, 3049F, 3050F, 39469-2, 49132-4, 55440-2, 80061, 83700, 83701, 83704, 83715, 83716, 83721

Microalbumin test

11218-5, 12842-1, 13705-9, 13801-6, 14585-4, 14956-7, 14957-5, 14958-3, 14959-1, 1753-3, 1754-1, 1755-8, 1757-4, 18373-1, 20454-5, 20621-9, 21059-1, 21482-5, 24356-8, 24357-6, 26801-1, 27298-9, 2887-8, 2888-6, 2889-4, 2890-2, 30000-4, 30001-2, 30003-8, 3062F, 32209-9, 32294-1, 32551-4, 34366-5, 35663-4, 40486-3, 40662-9, 40663-7, 43605-5, 43606-3, 43607-1, 44292-1, 47558-2, 49023-5, 50949-7, 53121-0, 53525-2, 53530-2, 53531-0, 53532-8, 56553-1, 57369-1, 57735-3, 5804-0, 58448-2, 58992-9, 59159-4, 82042, 82043, 82044, 83518, 84155, 84156, 84160, 84165, 84166, 9318-7

Retinal eye exam

14.1, 14.3, 14.4, 14.5, 14.9, 95.02, 95.03, 95.04, 95.11, 95.12, 95.16, 14.2, 2022F, 2024F, 2026F, 3072F, 67028, 67030, 67031, 67036, 67038, 67039, 67040, 67041, 67042, 67043, 67101, 67105, 67107, 67108, 67110, 67112, 67113, 67121, 67141, 67145, 67208, 67210, 67218, 67220, 67221, 67227, 67228, 92002, 92004, 92012, 92014, 92018, 92019, 92134, 92225, 92226, 92227, 92228, 92230, 92235, 92240, 92250, 92260, 92287, 99203, 99204, 99205, 99213, 99214, 99215, 99242, 99243, 99244, 99245, S0620, S0621, S0625, S3000, V72.0

Abbreviations: CPT, Current Procedural Terminology; HCPCS, Healthcare Common Procedure Coding System; LOINC, Logical Observation Identifiers Names and Codes; ICD-9-CM, International Classification of Diseases, 9th Revision, Clinical Modification; DRG, Diagnosis Related Group; AHFS, American Hospital Formulary Service

Page 8

eTable 4. ICD-9 codes categorized as acute complications of diabetes.

ICD-9 Code ICD-9 Code Description

Indicates Recent Acute Complication of Diabetes or Related Comorbidities

when Coded in:

Outpatient Setting

Emergency Department

Setting250.0

2 Diabetes mellitus without mention of complication, type II or unspecified type, uncontrolled 0 1

250.03 Diabetes mellitus without mention of complication, type I [juvenile type], uncontrolled 0 1

250.10 Diabetes with ketoacidosis, type II or unspecified type, not stated as uncontrolled 1 1

250.11 Diabetes with ketoacidosis, type I [juvenile type], not stated as uncontrolled 1 1

250.12 Diabetes with ketoacidosis, type II or unspecified type, uncontrolled 1 1

250.13 Diabetes with ketoacidosis, type I [juvenile type], uncontrolled 1 1

250.20 Diabetes with hyperosmolarity, type II or unspecified type, not stated as uncontrolled 1 1

250.21 Diabetes with hyperosmolarity, type I [juvenile type], not stated as uncontrolled 1 1

250.22 Diabetes with hyperosmolarity, type II or unspecified type, uncontrolled 1 1

250.23 Diabetes with hyperosmolarity, type I [juvenile type], uncontrolled 1 1

250.30 Diabetes with other coma, type II or unspecified type, not stated as uncontrolled 1 1

250.31 Diabetes with other coma, type I [juvenile type], not stated as uncontrolled 1 1

250.32 Diabetes with other coma, type II or unspecified type, uncontrolled 1 1

250.33 Diabetes with other coma, type I [juvenile type], uncontrolled 1 1

250.40 Diabetes with renal manifestations, type II or unspecified type, not stated as uncontrolled 0 1

250.42 Diabetes with renal manifestations, type II or unspecified type, uncontrolled 0 1

250.43 Diabetes with renal manifestations, type I [juvenile type], uncontrolled 0 1

250.52 Diabetes with ophthalmic manifestations, type II or unspecified type, uncontrolled 0 1

250.53 Diabetes with ophthalmic manifestations, type I [juvenile type], uncontrolled 0 1

250.62 Diabetes with neurological manifestations, type II or unspecified type, uncontrolled 0 1

250.63 Diabetes with neurological manifestations, type I [juvenile type], uncontrolled 0 1

250.72 Diabetes with peripheral circulatory disorders, type II or unspecified type, uncontrolled 0 1

250.73 Diabetes with peripheral circulatory disorders, type I [juvenile type], uncontrolled 0 1

250.82 Diabetes with other specified manifestations, type II or unspecified type, uncontrolled 0 1

250.83 Diabetes with other specified manifestations, type I [juvenile type], uncontrolled 0 1

250.92 Diabetes with unspecified complication, type II or unspecified type, uncontrolled 0 1

250.9 Diabetes with unspecified complication, type I [juvenile type], uncontrolled 0 1

Page 9

3251.0 Hypoglycemic coma 1 1251.1 Other specified hypoglycemia 1 1251.2 Hypoglycemia, unspecified 1 1276.7x Hyperkalemia 1 1337.1x Peripheral autonomic neuropathy 0 1357.2x Polyneuropathy in diabetes 0 1365.x Glaucoma 0 1

380.1x Infective otitis externa 1 1401.x Essential hypertension 0 1405.x Secondary hypertension 0 1410.x Acute myocardial infarction 1 1411.x Other acute and subacute forms of ischemic heart disease 1 1413.x Angina pectoris 1 1426.x Conduction disorders 0 1427.x Cardiac dysrhythmias 0 1428.x Heart failure 0 1430.x Subarachnoid hemorrhage 1 1431.x Intracerebral hemorrhage 1 1432.x Oth & unspec intracranial hemorrhage 1 1433.x Occlusion and stenosis of precerebral arteries 1 1434.x Occlusion of cerebral arteries 1 1435.x Transient cerebral ischemia 1 1436.x Acute, but ill-defined, cerebrovascular disease 1 1441.x Aortic aneurysm and dissection 1 1444.x Arterial embolism and thrombosis 1 1458.x Hypotension 1 1480.x Pneumonia due to virus 1 1481.x Pneumococcal pneumonia 1 1482.x Other bacterial pneumonia 1 1483.x Pneumonia due to mycoplasma, chlamydia, or other specified bacteria 1 1

484.x Pneumonia in whooping cough, anthrax, aspergillosis, other systemic mycoses, other infectious diseases classified elsewhere 1 1

485 Bronchopneumonia organism unspecified 1 1486 Pneumonia organism unspecified 1 1

487.x Influenza 1 1580.x Acute glomerulonephritis 1 1584.x Acute kidney failure 1 1590.x Infections of kidney 1 1595.x Cystitis 1 1

599.0x UTI site not specified 1 1681.x Cellulitis & abscess of finger & toe 1 1682.x Other cellulitis and abscess 1 1707.x Chronic ulcer of skin 0 1

729.2x Neuralgia, neuritis, and radiculitis, unspecified 0 1730.1x Chronic osteomyelitis ankle & foot 0 1785.4x Gangrene 1 1790.7x Bacteremia 1 1841.0 Lower limb amputation nos 1 1841.1 Amputation of toe 1 1841.2 Amputation through foot 1 1841.3 Disarticulation of ankle 1 1841.4 Amp ank thru malleoli tibia & fibula 1 1841.5 Other amputation below knee 1 1841.6 Disarticulation of knee 1 1841.7 Amputation above knee 1 1841.8 Disarticulation of hip 1 1841.9 Abdominopelvic amputation 1 0885.x Traumatic amputation of thumb 1 1886.x Traumatic amputation other finger 1 1

Page 10

887.x Traumatic amputation of arm & hand 1 1895.x Traumatic amputation of toe 1 1896.x Traumatic amputation of foot 1 1897.x Traumatic amputation of leg 1 1962.3 Poisoning by insulins and antidiabetic agents 1 1

Page 11

eTable 5. Frequency of acute diabetes complications.a

Acute complication in the day following the

period of interest?Patient Has a Gap in

Careb (n=39,590)

Matched Patient has no Contemporaneous Gap

in Care (n=39,590)No 39,410 39,566Yes 179 27

aSee eFigure 1 for definition of acute diabetes complication bSee “Validation of acute diabetes complication visit measure” above for definition of “gap in care”

eTable 6. Frequency of hospitalizations subsequent to acute diabetes complications visits a versus other outpatient or emergency department visits.

Hospitalization after: Acute Complication Visitb Non-acute Complication

Visit Outpatient Visits (n=42,963) (n=6,603,486)

No 38,298 6,412,999Yes 4665 190,487

Emergency Department Visits (n=77,278) (n=239,047)

No 56,272 212,737Yes 26,310 21,006

aSee eFigure 1 for definition of acute diabetes complication

Page 12

eTable 7. Out-of-pocket spending among HDHP group members one year before and two years after a HDHP switch compared to contemporaneous control group members.

Annual Amounts, per Member1Change in HDHP vs Control Group,

Year 1 vs Baseline1Change in HDHP vs Control Group,

Year 2 vs Baseline1

HDHP Group Control Group Relative, % Relative, %Baseline Year 1 Year 2 Baseline Year 1 Year 2 Estimate (95% CI) Estimate (95% CI)

Overall Cohort (n=12,084 per group)Total Out-of-pocket 1439.61 1918.77 2105.47 1425.70 1494.30 1634.35 27.2% (23.1%, 31.3%) 27.6% (23.4%, 31.8%)

Medical 695.86 1133.33 1312.58 665.85 726.05 854.98 49.4% (40.3%, 58.4%) 46.9% (38.0%, 55.8%)Prescription 749.13 848.45 877.25 757.73 795.87 824.72 7.8% (6.3%, 9.3%) 7.6% (5.8%, 9.4%)

Low-morbidity2 (n=7956 per group) Total Out-of-pocket 1007.12 1536.28 1720.16 1028.30 1212.27 1349.91 29.4% (25.4%, 33.4%) 30.1% (25.8%, 34.4%)


High-morbidity3 (n=3640 per group) Total Out-of-pocket 2336.55 2589.38 2755.35 2266.08 2018.45 2119.38 24.4% (19.5%, 29.3%) 26.1% (20.2%, 32.0%)


High-income4 (n=4555 per group) Total Out-of-pocket 1473.87 1950.58 2144.44 1484.31 1556.89 1719.37 26.2% (21.7%, 30.6%) 25.6% (20.8%, 30.4%)


Low-income5 (n=4121 per group) Total Out-of-pocket 1337.77 1864.98 2028.47 1302.65 1411.38 1573.80 28.7% (22.8%, 34.6%) 25.5% (19.7%, 31.3%)


Health Savings Account HDHP Members vs Matched Controls6 (n=1899 per group)

Total Out-of-pocket 1394.78 2175.92 2233.24 1369.54 1436.76 1531.01 48.7% (39.7%, 57.7%) 43.2% (34.2%, 52.2%)Medical 633.05 1145.57 1229.58 639.01 688.98 771.69 67.8% (47.9%, 87.8%) 60.8% (41.9%, 79.8%)Prescription 771.40 1070.83 1047.64 744.07 764.80 789.96 32.3% (26.2%, 38.4%) 26.3% (19.7%, 32.9%)

Higher-income7 (n=7800 per group)Total Out-of-pocket 1470.06 1933.83 2140.78 1489.91 1514.32 1675.31 29.4% (24.0%, 34.8%) 29.5% (24.0%, 35.1%)


Lower-income8 (n=7310 per group)Total Out-of-pocket 1391.24 1887.28 2051.94 1345.58 1433.95 1580.73 27.3% (23.2%, 31.4%) 25.5% (21.0%, 30.1%)


Lowest-income9 (n=1327 per group) Total Out-of-pocket 1231.90 1836.97 1968.69 1210.23 1353.17 1458.80 33.4% (23.0%, 43.7%) 32.6% (20.9%, 44.3%)


White Neighborhood Resident10 (n=8337 per group)

Page 13


Non-white Neighborhood Resident11 (n=3588 per group )


Abbreviations: HDHP, high-deductible health plan; CI, confidence interval. 1All rates and changes estimated using the STATA margins and nlcom commands, controlling for (as appropriate) age, gender, race/ethnicity, education level, poverty level, US region, ACG score, employer size, and index date. 2Adjusted Clinical Groups (see manuscript) score of <2. 3Adjusted Clinical Groups (see manuscript) score of ≥3. 4Living in neighborhoods with below-poverty levels of less than 5%. 5Living in neighborhoods with below-poverty levels of 10% or greater. 6Health Savings Accounts allow pre-tax contributions from employers or members, funds that can be used to pay for qualified medical expenses. 7Living in neighborhoods with below-poverty levels of less than 10%. 8Living in neighborhoods with below-poverty levels of 5% or greater. 9Living in neighborhoods with below-poverty levels of 20% or greater. 10Living in neighborhoods with at least 75% of residents having white race. 11Living in neighborhoods with fewer than 75% or residents having white race.

eTable 8. Mean out-of-pocket expenditure per test or visit among HDHP group members before and after a HDHP switch compared to contemporaneous control group members.

Mean Out-of-pocket Expenditure per Test or Visit, $*HDHP Group Control Group

Baseline Year 1 Year 2 Baseline Year 1 Year 2Overall Cohort

Primary Care Visit 15.4 23.3 26.8 14.5 14.2 14.7Specialist Visit 23.3 41.6 41.8 20.2 19.5 20.6Hemoglobin A1c Test 1.2 2.9 3.3 1.1 1.2 1.0LDL-C Test 1.4 3.2 4.8 1.4 1.3 1.3Microalbumin Test 1.3 2.8 2.2 0.6 0.7 1.0Retinal Eye Exam 24.8 39.5 43.7 22.4 23.6 24.7

Abbreviation: HDHP, high-deductible health plan. *Calculated during the first month of each enrollment year when members are least likely to have exceeded their annual deductible level.

Page 14

eTable 9. Monthly trends in primary outcomes during the baseline year comparing HDHP and control groups.Baseline Trend ofDifferenced Series P-value

Overall CohortPrimary Care Visits per Member per Month * (>0.2)Specialist Visits per Member per Month * (>0.2)

Low-morbidity2

Primary Care Visits per Member per Month -0.00032 (<0.001)Specialist Visits per Member per Month * (>0.2)

High-morbidity3

Primary Care Visits per Member per Month -0.00108 (0.089)Specialist Visits per Member per Month * (>0.2)

High-income4

Primary Care Visits per Member per Month * (>0.2)Specialist Visits per Member per Month * (>0.2)

Low-income5


Health Savings Account HDHP Members vs Matched Controls6


Higher-income7


Lower-income8


Lowest-income9


White Neighborhood Resident10


Non-white Neighborhood Resident11


Abbreviations: HDHP, high-deductible health plan. *P-value > 0.2 and thus excluded from regression model. 1Estimates derived from SAS proc autoreg applied to the differenced series (HDHP minus control), adjusting standard errors for autocorrelation. Terms with a p-value over 0.2 were excluded in final models. 2Adjusted Clinical Groups (see manuscript) score of <2. 3Adjusted Clinical Groups (see manuscript) score of ≥3. 4Living in neighborhoods with below-poverty levels of less than 5%. 5Living in neighborhoods with below-poverty levels of 10% or greater. 6Health Savings Accounts allow pre-tax contributions from employers or members, funds that can be used to pay for qualified medical expenses. 7Living in neighborhoods with below-poverty levels of less than 10%. 8Living in neighborhoods with below-poverty levels of 5% or greater. 9Living in neighborhoods with below-poverty levels of 20% or greater. 10Living in neighborhoods with at least 75% of residents having white race. 11Living in neighborhoods with fewer than 75% or residents having white race.

Page 15

eTable 10. Hazard ratios of time to first outpatient and emergency department preventable diabetes acute complication visit* in the baseline and follow-up periods.

Baseline Year Follow-up PeriodHazard Ratio,1

HDHP vs Control (95% CI)Hazard Ratio,1

HDHP vs Control (95% CI)First Outpatient Complication Visit

Overall Cohort 1.01 (0.93,1.09) 0.94 (0.88, 0.99)

Low-morbidity2 1.05 (0.92,1.19) 0.98 (0.91, 1.06)

High-morbidity3 0.97 (0.88,1.08) 0.89 (0.82, 0.97)

High-income4 1.00 (0.88,1.15) 0.95 (0.86, 1.04)

Low-income5 0.98 (0.86,1.11) 0.89 (0.81, 0.98)

HSA HDHP Members vs Matched Controls6 1.16 (0.95,1.41) 0.95 (0.81, 1.10)

Higher-income7 1.01 (0.92,1.12) 0.97 (0.90, 1.04)

Lower-income8 0.96 (0.87,1.06) 0.91 (0.85, 0.98)

Lowest-income9 1.11 (0.87,1.41) 0.90 (0.75, 1.07)

White Neighborhood Resident10 1.01 (0.92,1.10) 0.95 (0.89, 1.02)

Non-white Neighborhood Resident11 1.03 (0.89,1.19) 0.92 (0.82, 1.03)

First Emergency Department Complication Visit

Overall Cohort 0.93 (0.82,1.06) 1.00 (0.91, 1.10)

Low-morbidity2 0.85 (0.66,1.11) 0.94 (0.83, 1.08)

High-morbidity3 1.02 (0.88,1.19) 0.95 (0.84, 1.09)

High-income4 1.00 (0.79,1.26) 0.89 (0.75, 1.04)

Low-income5 0.92 (0.75,1.14) 1.14 (0.98, 1.33)

HSa HDHP Members vs Matched Controls6 1.05 (0.74,1.48) 1.17 (0.92, 1.51)

Higher-income7 0.93 (0.78,1.10) 0.91 (0.80, 1.02)

Page 16

Lower-income8 0.88 (0.75,1.03) 1.04 (0.93, 1.17)

Lowest-income9 0.86 (0.59,1.25) 1.07 (0.82, 1.39)

White Neighborhood Resident10 1.05 (0.89,1.23) 0.92 (0.82, 1.02)

Non-white Neighborhood Resident11 0.99 (0.77,1.27) 1.08 (0.91, 1.28)

Abbreviations: HDHP, high-deductible health plan; CI, confidence interval; HSA, health savings account. *See manuscript for definition. 1Hazard ratios estimated using Cox proportional hazard regression adjusted for (as appropriate) age, gender, race/ethnicity, education level, poverty level, US region, ACG score, employer size, and index date. 2Adjusted Clinical Groups (see manuscript) score of <2. 3Adjusted Clinical Groups (see manuscript) score of ≥3. 4Living in neighborhoods with below-poverty levels of less than 5%. 5Living in neighborhoods with below-poverty levels of 10% or greater. 6Health Savings Accounts allow pre-tax contributions from employers or members, funds that can be used to pay for qualified medical expenses. 7Living in neighborhoods with below-poverty levels of less than 10%. 8Living in neighborhoods with below-poverty levels of 5% or greater. 9Living in neighborhoods with below-poverty levels of 20% or greater. 10Living in neighborhoods with at least 75% of residents having white race. 11Living in neighborhoods with fewer than 75% or residents having white race.

eTable 11. Rates of and changes in in emergency department complication visits,* and 7-day emergency department complication episode expenditures before and after a HDHP switch compared to contemporaneous control group members.

Mean Annual Rates or Mean Expenditures1 Relative Change in HDHP vs Control Group, Follow-up

vs Baseline1HDHP Group Control GroupBaseline Follow-up Baseline Follow-up Estimate (95% CI)

Overall CohortMean Acute Complication Visits per 1000 Members 32.1 36.2 35.7 34.6 8.0% (4.6%, 11.4%)Mean Acute Complication Episode Expenditures per Member, $ 301.5 358.3 301.5 321.0 5.6% (3.8%, 7.3%)

Low-morbidity2

Mean Acute Complication Visits per 1000 Members 9.9 25.1 11.0 25.0 2.3% (-1.0%, 5.6%)Mean Acute Complication Episode Expenditures per Member, $ 21.9 209.0 21.9 229.1 -4.2% (-5.8%, -2.6%)

High-morbidity3

Mean Acute Complication Visits per 1000 Members 79.4 59.1 70.4 57.3 -5.2% (-9.4%, -1.0%)Mean Acute Complication Episode Expenditures per Member, $ 886.5 617.7 886.5 493.0 12.1% (7.2%, 17.0%)

High-income4

Mean Acute Complication Visits per 1000 Members 24.5 28.3 22.7 31.6 -7.3% (-9.7%, -4.8%)Mean Acute Complication Episode Expenditures per Member, $ 175.7 232.7 213.3 270.3 ND - -

Low-income5

Mean Acute Complication Visits per 1000 Members 36.9 44.3 41.4 35.0 21.7% (14.5%, 28.9%)Mean Acute Complication Episode Expenditures per Member, $ 321.9 439.0 321.9 366.2 9.5% (6.5%, 12.5%)

Health Savings Account HDHP Members vs Matched Controls 6


Higher-income7

Page 17

Mean Acute Complication Visits per 1000 Members 28.8 32.5 30.3 33.8 -2.0% (-3.1%, -0.8%)Mean Acute Complication Episode Expenditures per Member, $ 267.1 315.0 292.6 315.0 ND - -

Lower-income8

Mean Acute Complication Visits per 1000 Members 32.6 40.3 36.1 36.2 11.2% (6.9%, 15.4%)Mean Acute Complication Episode Expenditures per Member, $ 329.2 404.4 272.6 347.9 ND - -

Lowest-income9



Mean Acute Complication Visits per 1000 Members 30.1 34.7 27.4 37.2 -6.3% (-8.6%, -4.0%)Mean Acute Complication Episode Expenditures per Member, $ 305.6 348.7 224.6 337.1 -7.9% (-11.8%, -4.1%)



Abbreviations: HDHP, high-deductible health plan; CI, confidence interval; ND, difference not detected. *See manuscript for definition. 1All rates and changes estimated using the STATA margins and nlcom commands and adjust for any differences in baseline trends, additionally controlling for (as appropriate) age, gender, race/ethnicity, education level, poverty level (as appropriate), US region, ACG (as appropriate) score, employer size, and index date. 2Adjusted Clinical Groups (see manuscript) score of <2. 3Adjusted Clinical Groups (see manuscript) score of ≥3. 4Living in neighborhoods with below-poverty levels of less than 5%. 5Living in neighborhoods with below-poverty levels of 10% or greater. 6Health Savings Accounts allow pre-tax contributions from employers or members, funds that can be used to pay for qualified medical expenses. 7Living in neighborhoods with below-poverty levels of less than 10%. 8Living in neighborhoods with below-poverty levels of 5% or greater. 9Living in neighborhoods with below-poverty levels of 20% or greater. 10Living in neighborhoods with at least 75% of residents having white race. 11Living in neighborhoods with fewer than 75% or residents having white race.

Page 18

eTable 12. High-priority outpatient visits among key HDHP subgroup members before and after a HDHP switch compared to contemporaneous control group members.Mean Annual Rates, per Member1 Relative Change in HDHP vs

Control Group, Year 1 vs Baseline1Relative Change in HDHP vs

Control Group, Year 2 vs Baseline1HDHP Group Control GroupBaseline Year 1 Year 2 Baseline Year 1 Year 2 Estimate (95% CI) Estimate (95% CI)

Higher-income2

Primary Care 2.01 1.86 1.84 2.04 1.88 1.90 0.4% (-3.2%, 4.0%) -1.3% (-5.0%, 2.5%)Specialist 1.49 1.49 1.58 1.49 1.59 1.73 -5.9% (-10.7%, -1.2%) -8.7% (-13.8%, -3.6%)

Lower-income3

Primary Care 2.18 2.03 2.02 2.18 2.08 2.10 -2.0% (-5.4%, 1.3%) -3.5% (-7.1%, 0.1%)Specialist 1.35 1.43 1.50 1.34 1.49 1.60 -5.4% (-10.9%, 0.0%) -6.9% (-12.9%, -0.9%)

Lowest-income4

Primary Care 2.34 2.14 2.12 2.35 2.16 2.12 -0.3% (-8.0%, 7.3%) 0.4% (-7.9%, 8.7%)Specialist 1.17 1.27 1.36 1.17 1.37 1.40 -8.0% (-21.4%, 5.5%) -3.2% (-18.3%, 12.0%)


Primary Care 2.04 1.90 1.91 2.05 1.92 1.94 -0.5% (-3.9%, 2.8%) -0.8% (-4.4%, 2.8%)Specialist 1.46 1.50 1.57 1.48 1.61 1.72 -6.3% (-11.0%, -1.7%) -8.1% (-13.2%, -2.9%)


Primary Care 2.19 2.00 1.95 2.22 2.08 2.09 -2.4% (-7.3%, 2.4%) -4.9% (-9.8%, 0.1%)Specialist 1.31 1.32 1.40 1.31 1.39 1.48 -4.2% (-12.3%, 3.8%) -5.0% (-13.9%, 3.9%)

Abbreviations: HDHP, high-deductible health plan; CI, confidence interval. 1All rates and changes estimated using the STATA margins and nlcom commands and adjust for any differences in baseline trends, additionally controlling for (as appropriate) age, gender, race/ethnicity, education level, poverty level, US region, ACG score, employer size, and index date. 2Living in neighborhoods with below-poverty levels of less than 10%. 3Living in neighborhoods with below-poverty levels of 5% or greater. 4Living in neighborhoods with below-poverty levels of 20% or greater. 5Living in neighborhoods with at least 75% of residents having white race. 6Living in neighborhoods with fewer than 75% or residents having white race.

Page 19

Page 20

Page 21

eTable 13. Rates of and changes in disease monitoring measures among key HDHP subgroup members before and after a HDHP switch compared to contemporaneous control group members.

Mean Annual Rates, per Member1Relative Change in HDHP vs Control

Group1 Year 1 vs Baseline1Relative Change in HDHP vs Control

Group1 Year 2 vs Baseline1

HDHP Group Control GroupBaseline Year 1 Year 2 Baseline Year 1 Year 2 Estimate (95% CI) Estimate (95% CI)

Higher-income2

Primary Care Visit 0.88 0.87 0.80 0.89 0.87 0.81 0.5% (-0.9%, 2.0%) -0.7% (-2.2%, 0.7%)Hemoglobin A1c Test 0.71 0.68 0.64 0.72 0.69 0.65 0.3% (-1.8%, 2.4%) 0.1% (-2.1%, 2.3%)LDL-C Test 0.67 0.66 0.62 0.68 0.68 0.64 -1.0% (-3.4%, 1.4%) -2.3% (-4.7%, 0.2%)Microalbumin Test 0.35 0.36 0.36 0.35 0.35 0.36 2.4% (-2.6%, 7.3%) 0.0% (-4.8%, 4.8%)Retinal Eye Exam 0.31 0.33 0.34 0.32 0.34 0.35 -0.3% (-5.5%, 4.8%) -2.5% (-7.3%, 2.4%)

Lower-income3

Primary Care Visit 0.89 0.87 0.80 0.90 0.88 0.81 0.2% (-1.3%, 1.6%) -0.7% (-2.2%, 0.7%)Hemoglobin A1c Test 0.69 0.66 0.62 0.70 0.67 0.63 0.0% (-2.3%, 2.3%) -0.9% (-3.3%, 1.5%)LDL-C Test 0.65 0.64 0.60 0.66 0.65 0.61 0.3% (-2.4%, 3.0%) -0.7% (-3.4%, 2.0%)Microalbumin Test 0.32 0.33 0.33 0.32 0.33 0.32 -0.2% (-5.6%, 5.2%) 1.2% (-4.3%, 6.8%)Retinal Eye Exam 0.29 0.29 0.30 0.30 0.31 0.32 -1.7% (-7.4%, 4.0%) -1.6% (-7.2%, 4.0%)

Lowest-income4

Primary Care Visit 0.89 0.88 0.80 0.90 0.88 0.81 0.8% (-2.6%, 4.1%) -0.3% (-3.6%, 3.1%)Hemoglobin A1c Test 0.67 0.66 0.61 0.69 0.65 0.61 4.2% (-1.5%, 10.0%) 1.0% (-4.9%, 6.8%)LDL-C Test 0.63 0.63 0.59 0.64 0.61 0.59 5.8% (-1.2%, 12.8%) 1.5% (-5.2%, 8.3%)Microalbumin Test 0.31 0.33 0.33 0.33 0.30 0.32 13.9% (-0.8%, 28.5%) 8.8% (-5.2%, 22.7%)Retinal Eye Exam 0.27 0.28 0.28 0.29 0.29 0.29 5.6% (-9.4%, 20.7%) 5.9% (-8.8%, 20.5%)


Primary Care Visit 0.89 0.87 0.81 0.89 0.88 0.81 0.0% (-1.3%, 1.3%) -0.5% (-1.9%, 0.9%)Hemoglobin A1c Test 0.70 0.67 0.62 0.71 0.69 0.64 -0.8% (-2.9%, 1.2%) -0.8% (-2.9%, 1.4%)LDL-C Test 0.66 0.64 0.60 0.67 0.67 0.62 -2.1% (-4.4%, 0.3%) -1.4% (-3.8%, 1.1%)Microalbumin Test 0.34 0.34 0.34 0.34 0.35 0.35 0.0% (-4.8%, 4.8%) -1.7% (-6.4%, 3.0%)Retinal Eye Exam 0.31 0.32 0.33 0.33 0.34 0.34 1.6% (-3.5%, 6.7%) 3.9% (-1.2%, 8.9%)


Primary Care Visit 0.87 0.86 0.78 0.88 0.86 0.80 1.0% (-1.2%, 3.1%) -1.5% (-3.7%, 0.7%)Hemoglobin A1c Test 0.70 0.68 0.63 0.72 0.69 0.65 1.6% (-1.8%, 4.9%) 0.5% (-2.9%, 3.9%)LDL-C Test 0.67 0.66 0.62 0.69 0.67 0.64 2.5% (-1.3%, 6.4%) 0.4% (-3.4%, 4.3%)Microalbumin Test 0.33 0.35 0.34 0.33 0.35 0.34 -0.4% (-7.9%, 7.1%) 0.0% (-7.5%, 7.6%)Retinal Eye Exam 0.27 0.28 0.29 0.28 0.31 0.31 -8.4% (-16.3%, -0.5%) -6.8% (-14.7%, 1.0%)

Abbreviations: HDHP, high-deductible health plan; CI, confidence interval. 1All rates and changes estimated using the STATA margins and nlcom commands and adjusted for (as appropriate) age, gender, race/ethnicity, education level, poverty level, US region, ACG score, employer size, and index date. 2Living in neighborhoods with below-poverty levels of less than 10%. 3Living in neighborhoods with below-poverty levels of 5% or greater. 4Living in neighborhoods with below-poverty levels of 20% or greater. 5Living in neighborhoods with at least 75% of residents having white race. 6Living in neighborhoods with fewer than 75% or residents having white race.

eTable 14. Sensitivity analysis among members age 18-64: rates of and changes in high-priority outpatient visits among HDHP group members before and after a HDHP switch compared to contemporaneous control group members.

Mean Annual Rates, per Member1 Relative Change in HDHP vs Control Group per Member, Year 1 vs

Baseline1

Relative Change in HDHP vs Control Group per Member, Year 2 vs

Baseline1HDHP Group Control GroupBaseline Year 1 Year 2 Baseline Year 1 Year 2 Estimate (95% CI) Estimate (95% CI)

Overall Cohort Primary Care 2.10 1.93 1.93 2.13 1.99 2.01 -1.1% (1.6%, -3.9%) -2.5% (0.4%, -5.4%)Specialist 1.43 1.45 1.52 1.41 1.52 1.65 -5.7% (-1.6%, -9.8%) -8.8% (-4.5%, -13.2%)

Low-morbidity2 Primary Care 1.84 1.79 1.77 1.85 1.85 1.87 -3.1% (0.2%, -6.5%) -4.8% (-1.4%, -8.3%)Specialist 0.81 0.98 1.10 0.80 1.03 1.19 -5.8% (0.1%, -11.7%) -7.7% (-1.2%, -14.2%)

High-morbidity3 Primary Care 2.68 2.28 2.27 2.72 2.28 2.31 1.4% (6.6%, -3.8%) 0.0% (5.5%, -5.6%)Specialist 2.71 2.29 2.22 2.63 2.36 2.40 -6.1% (-0.3%, -11.8%) -10.5% (-4.6%, -16.3%)

High-income4 Primary Care 1.94 1.79 1.78 1.99 1.83 1.85 0.4% (5.3%, -4.4%) -1.0% (4.2%, -6.2%)Specialist 1.51 1.49 1.57 1.47 1.58 1.73 -8.4% (-2.3%, -14.6%) -12.0% (-5.6%, -18.4%)

Low-income5 Primary Care 2.25 2.09 2.07 2.26 2.17 2.15 -3.1% (1.3%, -7.5%) -2.8% (1.9%, -7.5%)Specialist 1.26 1.35 1.41 1.21 1.36 1.46 -5.2% (2.3%, -12.8%) -7.2% (1.1%, -15.6%)

Health Savings Account HDHP Members vs Matched Controls6 Primary Care 2.03 1.90 1.89 1.98 1.93 1.89 -3.8% (3.1%, -10.6%) -2.2% (5.0%, -9.4%)Specialist 1.63 1.64 1.79 1.48 1.61 1.67 -7.5% (2.4%, -17.4%) -2.4% (9.8%, -14.6%)

Abbreviations: HDHP, high-deductible health plan; CI, confidence interval. 1All rates and changes estimated using the STATA margins and nlcom commands and adjust for any differences in baseline trends, additionally controlling for (as appropriate) age, gender, race/ethnicity, education level, poverty level, US region, ACG score, employer size, and index date. 2Adjusted Clinical Groups (see text) score of <2. 3 Adjusted Clinical Groups (see text) score of ≥3. 4Living in neighborhoods with below-poverty levels of less than 5%. 5Living in neighborhoods with below-poverty levels of 10% or greater. 6Health Savings Accounts allow pre-tax contributions from employers or members, funds that can be used to pay for qualified medical expenses.

Page 22

eTable 15. Sensitivity analysis among members age 18-64: rates of and changes in disease monitoring measures among key HDHP subgroup members before and after a HDHP switch compared to contemporaneous control group members.

Mean Annual Rates, per Member1

Relative Change in HDHP vs Control Group1 Year 1 vs

Baseline1

Relative Change in HDHP vs Control Group1 Year 2 vs

Baseline1

HDHP Group Control Group

Baseli

ne

Year

1

Year

2

Baseli

ne

Year

1

Year

2 Estimate (95% CI) Estimate (95% CI)

Overall CohortPrimary Care Visit 0.89 0.87 0.80 0.89 0.88 0.81 0.0% (-1.1%, 1.1%) -0.8% (-1.9%, 0.4%)Hemoglobin A1c Test 0.70 0.67 0.63 0.71 0.68 0.64 -0.3% (-2.1%, 1.4%) -1.0% (-2.8%, 0.8%)LDL-C Test 0.66 0.65 0.61 0.68 0.67 0.63 0.2% (-1.8%, 2.3%) -0.9% (-2.9%, 1.2%)Microalbumin Test 0.33 0.35 0.34 0.34 0.35 0.35 0.4% (-3.6%, 4.5%) 0.0% (-4.0%, 4.1%)Retinal Eye Exam 0.30 0.31 0.32 0.31 0.33 0.33 0.2% (-4.1%, 4.6%) 0.1% (-4.2%, 4.4%)

Low-morbidity2 Primary Care Visit 0.87 0.86 0.79 0.87 0.87 0.81 -0.5% (-2.0%, 1.0%) -1.8% (-3.3%, -0.3%)Hemoglobin A1c Test 0.69 0.67 0.63 0.70 0.69 0.65 -0.9% (-3.1%, 1.3%) -2.0% (-4.2%, 0.2%)LDL-C Test 0.65 0.65 0.61 0.67 0.67 0.64 0.6% (-1.9%, 3.2%) -1.2% (-3.8%, 1.3%)Microalbumin Test 0.33 0.35 0.34 0.33 0.35 0.35 -1.1% (-6.0%, 3.8%) -1.7% (-6.6%, 3.2%)Retinal Eye Exam 0.26 0.29 0.29 0.27 0.30 0.31 -0.5% (-6.4%, 5.4%) -1.6% (-7.2%, 4.0%)

High-morbidity3 Primary Care Visit 0.93 0.90 0.83 0.94 0.90 0.84 1.3% (-0.4%, 3.1%) 0.6% (-1.1%, 2.4%)Hemoglobin A1c Test 0.72 0.68 0.63 0.73 0.68 0.64 1.0% (-2.2%, 4.1%) -0.4% (-3.6%, 2.8%)LDL-C Test 0.69 0.66 0.62 0.70 0.68 0.64 -1.4% (-4.8%, 2.1%) -1.3% (-4.8%, 2.3%)Microalbumin Test 0.35 0.35 0.34 0.37 0.35 0.35 3.9% (-3.5%, 11.3%) 1.4% (-5.8%, 8.6%)Retinal Eye Exam 0.38 0.37 0.37 0.40 0.39 0.39 -1.8% (-8.3%, 4.7%) 1.8% (-4.9%, 8.4%)

High-income4 Primary Care Visit 0.88 0.86 0.80 0.89 0.86 0.81 1.1% (-0.8%, 2.9%) -1.1% (-3.0%, 0.8%)Hemoglobin A1c Test 0.72 0.69 0.65 0.73 0.70 0.66 -1.0% (-3.7%, 1.7%) -1.1% (-3.9%, 1.7%)LDL-C Test 0.68 0.67 0.63 0.70 0.69 0.65 0.3% (-2.9%, 3.4%) 0.0% (-3.2%, 3.2%)Microalbumin Test 0.36 0.38 0.36 0.36 0.37 0.38 1.7% (-4.6%, 8.0%) -2.2% (-8.3%, 3.9%)Retinal Eye Exam 0.32 0.34 0.35 0.34 0.36 0.36 2.1% (-4.7%, 8.9%) 2.9% (-3.6%, 9.4%)

Low-income5 Primary Care Visit 0.89 0.88 0.80 0.90 0.88 0.82 0.4% (-1.5%, 2.3%) -0.6% (-2.6%, 1.4%)Hemoglobin A1c Test 0.68 0.66 0.61 0.69 0.66 0.62 0.4% (-2.8%, 3.6%) -1.8% (-5.1%, 1.4%)LDL-C Test 0.63 0.64 0.59 0.64 0.63 0.60 2.9% (-0.9%, 6.7%) 0.4% (-3.4%, 4.1%)Microalbumin Test 0.30 0.32 0.31 0.31 0.32 0.32 3.2% (-4.4%, 10.9%) 2.4% (-5.3%, 10.1%)Retinal Eye Exam 0.27 0.29 0.29 0.28 0.31 0.30 -1.1% (-9.0%, 6.8%) -0.2% (-8.0%, 7.7%)

Health Savings Account HDHP Members vs Matched Controls6

Primary Care Visit 0.90 0.87 0.81 0.89 0.90 0.81 -3.6% (-6.3%, -1.0%) -1.3% (-4.1%, 1.5%)Hemoglobin A1c Test 0.70 0.68 0.64 0.70 0.67 0.63 1.4% (-3.0%, 5.9%) 1.1% (-3.6%, 5.8%)LDL-C Test 0.66 0.67 0.62 0.66 0.66 0.62 1.0% (-4.1%, 6.1%) 0.0% (-5.3%, 5.3%)Microalbumin Test 0.36 0.37 0.36 0.34 0.34 0.34 -0.5% (-10.3%, 9.3%) -2.3% (-12.0%, 7.4%)Retinal Eye Exam 0.32 0.36 0.36 0.30 0.32 0.34 3.8% (-7.1%, 14.7%) -0.2% (-10.6%, 10.3%)

Abbreviations: HDHP, high-deductible health plan; CI, confidence interval. 1All rates and changes estimated using the STATA margins and nlcom commands and adjust for any differences in baseline trends, additionally controlling for (as appropriate) age, gender, race/ethnicity, education level, poverty level (as appropriate), US region, ACG (as appropriate) score, employer size, and index date. 2Adjusted Clinical Groups (see text) score of <2. 3 Adjusted Clinical Groups (see text) score of ≥3. 4Living in neighborhoods with below-poverty levels of less than 5%. 5Living in neighborhoods with below-poverty levels of 10% or greater. 6Health Savings Accounts allow pre-tax contributions from employers or members, funds that can be used to pay for qualified medical expenses.

eTable 16. Sensitivity analysis among members age 18-64: hazard ratios of time to first outpatient and emergency department preventable diabetes acute complication visit* in the baseline and follow-up periods.

Baseline Year Follow-up PeriodHazard Ratio,1

HDHP vs Control (95% CI)Hazard Ratio,1

HDHP vs Control (95% CI)First Outpatient Complication Visit

Overall Cohort 0.98 (0.91, 1.06) 0.96 (0.90, 1.01)Low-morbidity2 1.01 (0.89, 1.14) 0.97 (0.89, 1.05)High-morbidity3 1.00 (0.90, 1.11) 0.89 (0.82, 0.98)High-income4 0.98 (0.85, 1.11) 0.98 (0.89, 1.08)Low-income5 0.99 (0.87, 1.13) 0.91 (0.83, 1.01)Health Savings Account HDHP Members vs Matched Controls6 1.03 (0.84, 1.25) 0.92 (0.79, 1.07)

First Emergency Department Complication VisitOverall Cohort 0.90 (0.79, 1.03) 0.99 (0.90, 1.08)Low-morbidity2 0.82 (0.63, 1.08) 1.07 (0.93, 1.22)High-morbidity3 1.06 (0.91, 1.24) 0.98 (0.86, 1.11)High-income4 0.90 (0.71, 1.15) 0.90 (0.77, 1.06)Low-income5 0.90 (0.73, 1.11) 1.12 (0.96, 1.30)Health Savings Account HDHP Members vs Matched Controls6 0.88 (0.63, 1.25) 0.93 (0.73, 1.18)Abbreviations: HDHP, high-deductible health plan; CI, confidence interval. *See manuscript for definition. 1Hazard ratios estimated using Cox proportional hazard regression adjusted for (as appropriate) age, gender, race/ethnicity, education level, poverty level, US region, ACG score, employer size, and index date. 2Adjusted Clinical Groups (see manuscript) score of <2. 3 Adjusted Clinical Groups (see manuscript) score of ≥3. 4Living in neighborhoods with below-poverty levels of less than 5%. 5Living in neighborhoods with below-poverty levels of 10% or greater. 6Health Savings Accounts allow pre-tax contributions from employers or members, funds that can be used to pay for qualified medical expenses.

Page 23

eTable 17. Sensitivity analysis among members age 18-64: rates of and changes in emergency department complication visits,* and 7-day emergency department complication episode expenditures before and after a HDHP switch compared to contemporaneous control group members.

Mean Annual Rates or Mean Expenditures1 Relative Change in HDHP vs Control Group, Follow-up

vs Baseline1HDHP Group Control GroupBaseline Follow-up Baseline Follow-up Estimate (95% CI)

Overall CohortMean Acute Complication Visits per 1000 Members 30.4 36.2 32.5 36.3 2.8% (1.4%, 4.2%)Mean Acute Complication Episode Expenditures per Member, $ 273.1 345.5 286.0 351.7 -0.8% (-2.0%, 0.3%)

Low-morbidity2


High-morbidity3

Mean Acute Complication Visits per 1000 Members 75.3 60.2 70.2 55.1 ND - -Mean Acute Complication Episode Expenditures per Member, $ 809.6 639.0 809.6 475.7 16.4% (9.8%, 23.0%)

High-income4

Mean Acute Complication Visits per 1000 Members 22.4 28.8 25.9 32.7 -5.7% (-7.4%, -3.9%)Mean Acute Complication Episode Expenditures per Member, $ 172.4 244.9 221.8 283.7 -6.5% (-10.1%, -3.0%)

Low-income5



Mean Acute Complication Visits per 1000 Members ** ** ** **Mean Acute Complication Episode Expenditures per Member, $ ** ** ** **

Abbreviations: HDHP, high-deductible health plan; CI, confidence interval; ND, difference not detected. *See manuscript for definition. **Statistical models would not converge due to rare events. 1All rates and changes estimated using the STATA margins and nlcom commands and adjust for any differences in baseline trends, additionally controlling for (as appropriate) age, gender, race/ethnicity, education level, poverty level (as appropriate), US region, ACG (as appropriate) score, employer size, and index date. 2Adjusted Clinical Groups (see manuscript) score of <2. 3Adjusted Clinical Groups (see manuscript) score of ≥3. 4Living in neighborhoods with below-poverty levels of less than 5%. 5Living in neighborhoods with below-poverty levels of 10% or greater. 6Health Savings Accounts allow pre-tax contributions from employers or members, funds that can be used to pay for qualified medical expenses.

Page 24

eTable 18. Sensitivity analysis restricting to the top 5 most common emergency department complication visits: changes in emergency department complication visits,* and 7-day emergency department complication episode expenditures before and after a HDHP switch compared to contemporaneous control group members.

Relative Change in HDHP vs Control Group, Follow-up

vs Baseline1

Estimate (95% CI)Overall Cohort

Mean Acute Complication Visits per 1000 Members 14.5% (10.0%, 19.0%)Mean Acute Complication Episode Expenditures per Member, $ 17.0% (11.8%, 22.3%)

Low-morbidity2

Mean Acute Complication Visits per 1000 Members 4.0% (0.6%, 7.3%)Mean Acute Complication Episode Expenditures per Member, $ -8.8% (-9.7%, -7.9%)

High-morbidity3

Mean Acute Complication Visits per 1000 Members -6.4% (-10.6%, -2.2%)Mean Acute Complication Episode Expenditures per Member, $ 36.4% (14.4%, 58.5%)

High-income4

Mean Acute Complication Visits per 1000 Members -2.1% (-3.6%, -0.6%)Mean Acute Complication Episode Expenditures per Member, $ 29.4% (13.4%, 45.4%)

Low-income5

Mean Acute Complication Visits per 1000 Members 37.5% (20.3%, 54.7%)Mean Acute Complication Episode Expenditures per Member, $ 65.6% (31.9%, 99.3%)


Mean Acute Complication Visits per 1000 Members 9.6% (-0.7%, 19.9%)Mean Acute Complication Episode Expenditures per Member, $ ND - -

Abbreviations: HDHP, high-deductible health plan; CI, confidence interval; ND, difference not detected. *See manuscript for definition. **Statistical models would not converge due to rare events. 1All rates and changes estimated using the STATA margins and nlcom commands and adjust for any differences in baseline trends, additionally controlling for (as appropriate) age, gender, race/ethnicity, education level, poverty level (as appropriate), US region, ACG (as appropriate) score, employer size, and index date. 2Adjusted Clinical Groups (see manuscript) score of <2. 3Adjusted Clinical Groups (see manuscript) score of ≥3. 4Living in neighborhoods with below-poverty levels of less than 5%. 5Living in neighborhoods with below-poverty levels of 10% or greater. 6Health Savings Accounts allow pre-tax contributions from employers or members, funds that can be used to pay for qualified medical expenses.

Page 25

Page 26

Page 27

Page 28

Page 29

Page 30

References

1. Cook EF, Goldman L. Performance of tests of significance based on stratification by a multivariate confounder score or by a propensity score. Journal of clinical epidemiology. 1989;42(4):317-324.

2. D'Agostino RB, Jr. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med. 1998;17(19):2265-2281.

3. Rubin DB. The design versus the analysis of observational studies for causal effects: parallels with the design of randomized trials. Stat Med. 2007;26(1):20-36.

4. Coca-Perraillon M. Local and Global Optimal Propensity Score Matching. 2007; http://www2.sas.com/proceedings/forum2007/185-2007.pdf. Accessed 19 December, 2009.

5. St.Clair T, Cook TD, Hallberg K. Examining the Internal Validity and Statistical Precision of the Comparative Interrupted Time Series Design by Comparison With a Randomized Experiment. American Journal of Evaluation. 2014;35(3):311-327.

6. Austin PC. Optimal caliper widths for propensity-score matching when estimating differences in means and differences in proportions in observational studies. Pharm Stat. 2010.

7. American Diabetes A. Economic costs of diabetes in the U.S. in 2012. Diabetes care. 2013;36(4):1033-1046.8. The Johns Hopkins ACG Case-Mix System Reference Manual, Version 7.0. Baltimore, MD: The Johns Hopkins University; 2005.9. Reid RJ, Roos NP, MacWilliam L, Frohlich N, Black C. Assessing population health care need using a claims-based ACG morbidity

measure: a validation analysis in the Province of Manitoba. Health Serv Res. 2002;37(5):1345-1364.10. U.S. Bureau of the Census. Geographical Areas Reference Manual, Washington, D.C., U.S. Bureau of the Census. 1994.11. Krieger N. Overcoming the absence of socioeconomic data in medical records: validation and application of a census-based methodology.

American Journal of Public Health. 1992;82(5):703-710.12. Krieger N, Chen JT, Waterman PD, Rehkopf DH, Subramanian SV. Race/ethnicity, gender, and monitoring socioeconomic gradients in

health: a comparison of area-based socioeconomic measures--the public health disparities geocoding project. American Journal of Public Health. 2003;93(10):1655-1671.

13. Trivedi AN, Zaslavsky AM, Schneider EC, Ayanian JZ. Relationship between quality of care and racial disparities in Medicare health plans. Jama. 2006;296(16):1998-2004.

14. Trivedi AN, Rakowski W, Ayanian JZ. Effect of cost sharing on screening mammography in medicare health plans. The New England journal of medicine. 2008;358(4):375-383.

15. Selby JV, Fireman BH, Swain BE. Effect of a copayment on use of the emergency department in a health maintenance organization. New England Journal of Medicine. 1996;334(10):635-641.

16. Ethnic Technologies. http://www.ethnictechnologies.com/index.html.17. Fiscella K, Fremont AM. Use of geocoding and surname analysis to estimate race and ethnicity. Health services research. 2006;41(4 Pt

1):1482-1500.18. Wagner AK, Soumerai SB, Zhang F, Ross-Degnan D. Segmented regression analysis of interrupted time series studies in medication use

research. Journal of clinical pharmacy and therapeutics. 2002;27(4):299-309.

Page 31

http://www.ethnictechnologies.com/index.html

http://www2.sas.com/proceedings/forum2007/185-2007.pdf

19. Liang KY, Zeger SL. Longitudinal Data Analysis Using Generalized Linear Models. Biometrika. 1986;73(1):13-22.20. Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics. 1986;42(1):121-130.21. Williams R. Using the margins command to estimate and interpret adjusted predictions and marginal effects. Stata Journal 2012;12(2):308-

331.22. McCullagh P, Nelder JA. Generalized linear models. 2nd ed. London ; New York: Chapman and Hall; 1989.23. Burr IW. Cumulative Frequency Functions. 1942:215-232.24. Kong E, Tong H, Xia Y. STATISTICAL MODELLING OF NONLINEAR LONG-TERM CUMULATIVE EFFECTS. Statistica Sinica.

2010;20(3):1097-1123.

Page 32

stacks.cdc.gov€¦ · Web vieweAppendix. I. eMethods. A. Study Group Construction and Deductible Imputation Algorithm. To determine employer deductible levels, we used a benefits

Documents