Predictive Model Development: Identifying high-risk patients in primary care attendance Financed with the help PI 06/1122 (FIS) of ISCIII Antonio Sarría Santamera Mª Auxiliadora Martín Martín Mª del Rocío Carmona Alférez Pilar Gallego Berciano AGENCY EVALUATION OF HEALTH TECHNOLOGIES AGENCY EVALUATION OF HEALTH TECHNOLOGIES INSTITUTO DE SALUD CARLOS III INSTITUTO DE SALUD CARLOS III
27
Embed
Predictive Model Development: Identifying high-risk patients in primary care attendance
Antonio Sarría Santamera Mª Auxiliadora Martín Martínez Mª del Rocío Carmona Alférez Pilar Gallego Berciano. Predictive Model Development: Identifying high-risk patients in primary care attendance. Financed with the help PI 06/1122 (FIS) of ISCIII. - PowerPoint PPT Presentation
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Predictive Model Development:
Identifying high-risk patients in primary care attendance
Financed with the help PI 06/1122 (FIS) of ISCIII
Antonio Sarría Santamera Mª Auxiliadora Martín Martínez Mª del Rocío Carmona AlférezPilar Gallego Berciano
AGENCY EVALUATION OF HEALTH TECHNOLOGIESAGENCY EVALUATION OF HEALTH TECHNOLOGIES
INSTITUTO DE SALUD CARLOS IIIINSTITUTO DE SALUD CARLOS III
Sandín Vázquez MConde Espejo Pde Bustos Guadaño MAsunsolo del Barco ARiesgo Fuertes RGarrido Elustondo SCabello Ballesteros MLEscortell Mayor MESanz Cuesta T
1. Total medical PC consultations (2006 and 2007).2. Total nursing PC consultations (2006 and 2007).3. Analytics4. Radiology tests5. Referrals to specialists
PRIMARY CARE TEAM VARIABLES
Organizational Characteristics and Health System Capacity
1. Location of the PCT2. Type of primary health care team3. Time schedule2. Average PC doctor workload3. Average PC nursing workload4. Percentage of patients aged ≥ 65 years of primary care team
Socioeconomic
1. Educational level of the Basic Health Zone (Population and Housing Census 2001 of the CM)2. Per capita Gross Disposable Income of the Basic Health Zone (Statistical Institute of the CM)
SOCIO-ECONOMIC DATA
Patient
Per Capita Gross Disposable Income in 2000 of the basic health area
SOCIO-ECONOMIC DATA
Patient
Per Capita Gross Disposable Income in 2000 of the basic health area
2.2.DEVELOPMENT OF RISK ADJUSTMENT MODELS: DEVELOPMENT OF RISK ADJUSTMENT MODELS: MULTILEVEL REGRESSIONMULTILEVEL REGRESSION
Characterized byCharacterized by: hierarchical clustering of variables: hierarchical clustering of variablesLEVEL 1: LEVEL 1: PATIENTSPATIENTSLEVEL 2: LEVEL 2: PC Teams PC Teams
DEPENDENT VARIABLEDEPENDENT VARIABLE: :
Total number of visits to PC physicians in 2007 Total number of visits to PC physicians in 2007
The variability in total of visits to PC physicians are due to differences The variability in total of visits to PC physicians are due to differences between patients and differences between PC Teamsbetween patients and differences between PC Teams
DATA ANALYSISDATA ANALYSIS
2.2.DEVELOPMENT OF RISK ADJUSTMENT MODELS: DEVELOPMENT OF RISK ADJUSTMENT MODELS: MULTILEVEL REGRESSIONMULTILEVEL REGRESSION
Characterized byCharacterized by: hierarchical clustering of variables: hierarchical clustering of variablesLEVEL 1: LEVEL 1: PATIENTSPATIENTSLEVEL 2: LEVEL 2: PC Teams PC Teams
DEPENDENT VARIABLEDEPENDENT VARIABLE: :
Total number of visits to PC physicians in 2007 Total number of visits to PC physicians in 2007
The variability in total of visits to PC physicians are due to differences The variability in total of visits to PC physicians are due to differences between patients and differences between PC Teamsbetween patients and differences between PC Teams
METHODOLOGY
MULTILEVEL MODEL:MULTILEVEL MODEL:
INDEPENDENT VARIABLES OF PATIENTINDEPENDENT VARIABLES OF PATIENT
MULTILEVEL MODEL:MULTILEVEL MODEL:
INDEPENDENT VARIABLES OF PATIENTINDEPENDENT VARIABLES OF PATIENT
METHODOLOGY
METHODOLOGYSociodemográficas1. Edad 2. Sexo 3. Problemas Sociales
Use of Health Service1. Total of visits to PC physicians in 2006 2. Total nursing consultations in 20063. Analytical
4. Radiology tests5. Referrals to other specialists
Sociodemográficas1. Edad 2. Sexo 3. Problemas Sociales
Use of Health Service1. Total of visits to PC physicians in 2006 2. Total nursing consultations in 20063. Analytical
4. Radiology tests5. Referrals to other specialists
Sociodemográficas1. Edad 2. Sexo 3. Problemas SocialesSociodemographic1. Age 2. Sex 3. Social Problems
INDEPENDENT VARIABLES OF PC TeamINDEPENDENT VARIABLES OF PC Team
MULTILEVEL MODEL :MULTILEVEL MODEL :
INDEPENDENT VARIABLES OF PC TeamINDEPENDENT VARIABLES OF PC Team
METHODOLOGY
Características y Capacidad Organizativa del Sistema Sanitario
1. Tipo de EAP: Rural/ Urbano 2. Turno del EAP: Mañana y Tarde/ Mañana o Tarde 3. PAMED 4. PAENF5. % de pacientes con edad ≥ 65 años del EAP
Organizational Characteristics and Health System Capacity
1. Type of PC Team: Rural / Urban2. Time Team's primary care: Morning and Evening / Morning or Afternoon3. Pressure care average PC physician PC Team4. Pressure half of nursing care of PC Team5. % of patients aged ≥ 65 years of PC Team
Socio-economic
Per capita disposable income 2000
METHODOLOGY
DATA ANALYSISDATA ANALYSIS
3.3.RISK ADJUSTED ATTENDANCE RATIORISK ADJUSTED ATTENDANCE RATIO
Estimated valuesEstimated values were obtained for each patient according were obtained for each patient according to the risk adjustment modelto the risk adjustment model..
DATA ANALYSISDATA ANALYSIS
3.3.RISK ADJUSTED ATTENDANCE RATIORISK ADJUSTED ATTENDANCE RATIO
Estimated valuesEstimated values were obtained for each patient according were obtained for each patient according to the risk adjustment modelto the risk adjustment model..
METHODOLOGY
DATA ANALYSISDATA ANALYSIS
3.3. RISK ADJUSTED ATTENDANCE RATIORISK ADJUSTED ATTENDANCE RATIO
For each patient we calculated the difference between For each patient we calculated the difference between the actual number of visits to PC physician in 2007 and the actual number of visits to PC physician in 2007 and the value estimated by the model:the value estimated by the model:
DATA ANALYSISDATA ANALYSIS
3.3. RISK ADJUSTED ATTENDANCE RATIORISK ADJUSTED ATTENDANCE RATIO
For each patient we calculated the difference between For each patient we calculated the difference between the actual number of visits to PC physician in 2007 and the actual number of visits to PC physician in 2007 and the value estimated by the model:the value estimated by the model:
Difference = OBSERVED - ESTIMATED
OBSERVED – ESTIMATED > 0 PATIENT OVER-USER
OBSERVED – ESTIMATED < 0 PATIENT INFRA-USER
METHODOLOGY
DATA ANALYSISDATA ANALYSIS
3.3. RISK ADJUSTED ATTENDANCE RATIO RISK ADJUSTED ATTENDANCE RATIO
This approach allowed to obtained the This approach allowed to obtained the observed/expeced number of visits observed/expeced number of visits for each PC Team. for each PC Team. The total number of visits to PC in 2007 of these The total number of visits to PC in 2007 of these patients and the estimated total number of patients and the estimated total number of consultations by the model.consultations by the model.
DATA ANALYSISDATA ANALYSIS
3.3. RISK ADJUSTED ATTENDANCE RATIO RISK ADJUSTED ATTENDANCE RATIO
This approach allowed to obtained the This approach allowed to obtained the observed/expeced number of visits observed/expeced number of visits for each PC Team. for each PC Team. The total number of visits to PC in 2007 of these The total number of visits to PC in 2007 of these patients and the estimated total number of patients and the estimated total number of consultations by the model.consultations by the model.
METHODOLOGY
PC Team
Total visits to PC Estimated number of visits
physicians PC physicians
1 185703 102218
2 219366 124761
3 207843 100874
4 616086 308746
5 173189 56313
DATA ANALYSISDATA ANALYSIS
3.3. RISK ADJUSTED ATTENDANCE RATIO RISK ADJUSTED ATTENDANCE RATIO
DATA ANALYSISDATA ANALYSIS
3.3. RISK ADJUSTED ATTENDANCE RATIO RISK ADJUSTED ATTENDANCE RATIO
METHODOLOGY
PC Team
Total visits to PC Estimated number of visits
RAAR = RATIOphysicians PC physicians
1 185703 102218 1.82
2 219366 124761 1.76
3 207843 100874 2.06
4 616086 308746 2.00
5 173189 56313 3.08
CHARACTERISTICS OF PATIENTS AND PC TEAM MEAN S.D. %
AGE 49.57 16.86
SEX: FEMALE 55.68
IMMIGRANT 10.54
TOTALVISITS TO PC PHYSICIANS 2007 14.59 19.71
TOTALVISITS TO PC PHYSICIANS 2006 13.15 17.27
TOTALNOURSING CONSULTATIONS 2007 2.64 4.20
TOTALNOURSING CONSULTATIONS 2006 2.65 4.10
ANALYTICAL 39.47
TOTAL OF PRESCRIBED DRUGS 5.16 5.93
REFERRAL 31.85
USER TYPE: ACTIVE / PENSIONER 68.56
PRESSURE CARE AVERAGE PC PHYSICIAN PC TEAM 39.21 5.01
PRESSURE HALF OF NURSING CARE OF PC TEAM 19.77 3.41
PERCENTAGE OF PATIENTS AGED ≥ 65 YEARS OF PC TEAM 13.52 5.56
Type of PC Team: Rural 91.97
Time Team's PC: Morning and Evening 80.71
PER CAPITA INCOME AVAILABLE BASIC HEALTH ZONE 10.044.49 2.406.55
RESULTS
VARIABLES Coeff. Z 95% CI
Age 0.01 104.35 0.01 0.01
Sex (Ref. Male) 0.12 72.52 0.11 0.12
Log Total visits to PC Physicians 2006 0.56 506.85 0.56 0.57
Log Total Noursing Consultations 2006 0.03 30.25 0.03 0.04
Other respiratory disease (Ref. No) 0.08 46.37 0.08 0.09
Skin / appendages (Ref. No) 0.03 15.70 0.02 0.03
Diabetes (Ref. No) 0.05 17.16 0.04 0.06
Obesity (Ref. No) 0.01 4.90 0.01 0.02
Lipid disorder (Ref. No) 0.05 21.74 0.04 0.05
Cancer (Ref. No) 0.03 10.07 0.02 0.04
Schizophrenia (Ref. No) 0.19 15.26 0.16 0.21
RESULTS
COEFFICIENT
STANDARD ERROR
LOG-LIKELIHOOD -1186575.2
Variability between PC Teams
0.29 0.0198
Variability between patients
0.78 0.0005
ICC 12.6% 0.0148
VARIABLES Coeff
.Z 95% CI
Type of PC Team (Ref. Rural) -0.21 -2.27 -0.39 -0.03
Pressure Half of Nursing Care of PC Team
0.022.57
0.00 0.04
3 Income categories (Ref. Low)
Intermediate -0.04 -0.49 -0.18 0.11
High/ Very High -0.18 -2.72 -0.31 -0.05
RESULTS
53% of patients visits their primary care physician more 53% of patients visits their primary care physician more than estimated according to their riskthan estimated according to their risk..
40% visits less primary care physicians than expected.40% visits less primary care physicians than expected.
RESULTS
PATIENT
OVER-USER 53%
INFRA-USER 40%
n= 1,017,002
There is a significant amount of visits to PC
physician which are not explained by the risk of
patients or characteristics of PC Teams.
There is a significant amount of visits to PC
physician which are not explained by the risk of
patients or characteristics of PC Teams.
Overall, the number of visits of over-users exceeds
that of infra-users.
Overall, the number of visits of over-users exceeds