Figures and tables in A Dynamic Microsimulation Model for Ageing and Health in England: The English Future Elderly Model

Abstract

Population ageing has the potential to disrupt every aspect of society, placing a great deal of stress in particular on our healthcare and welfare systems. To overcome these challenges will require effective policy interventions with effects that may not be truly seen for decades. Policy makers would therefore benefit from a tool that allows them to assess the long-term impact of their decisions on population health. We have developed such a tool by adapting the well-established Future Elderly Model (FEM) developed in the US to use the English Longitudinal Study of Ageing. The FEM is a Markov microsimulation model for people aged over 50 that generates input populations and transition probabilities using the English Longitudinal Study of Ageing, and can then project the population forward in time assessing the prevalence and incidence of a number of chronic diseases and related economic outputs. By modifying the input populations or transition probabilities, we can use the model to investigate counterfactual scenarios and assess the long-term effects of policy interventions on elderly health. In this paper we describe the model and its workings, provide evidence to validate its outputs, and outline possible applications.

This article contains 10 figures and 14 tables.

Figures

**Figure 1.** Structure of the FEM simulation.

**Figure 2.** Handover plots for Chronic diseases by sex.

**Figure 3.** ROC curves for key binary response variables with a 10-year simulation horizon, from 2002 - 2012.

**Figure 4.** Chronic Disease prevalence comparisons by age group between FEM and AgeUK almanac of chronic disease in 2014. AgeUK figures include 95% confidence intervals.

**Figure 5.** Smoking prevalence comparison by age group in 2018. Simulated data (starting in 2012) compared with prevalence statistics generated by the Action on Smoking and Health charity.

**Figure 6.** Average age at death comparison between baseline and no heart disease intervention. The shaded bands represent Monte Carlo confidence intervals.

**Figure 7.** Whole population (including smokers and non-smokers) survival curve compared between the baseline and no smoking scenarios.

**Figure 8.** Whole population prevalence of lung disease comparison between the baseline and no smoking scenarios, with Monte Carlo confidence intervals.

**Figure A1.** Handover plots for ADLs and IADLs.

**Figure A2.** Receiver Operating Characteristic curves.

Tables

Table 1. Variables tracked by the model

Domain	Variable
Health	Mortality, Alzheimers, Cancer, Dementia, Diabetes, Heart Disease, High Cholesterol, Hypertension, Lung Disease, Stroke
Risk Factors	BMI, Smoking Status, Alcohol Consumption, Exercise
Functional Limitations	Difficulties in Activities of Daily Living (ADLs), and Instrumental Activities of Daily Living (IADLs)
Economic	Employed, Unemployed, Retired/Disabled

Table 2. Sample attrition in the over 50s in ELSA. The sample was replenished in 2006, 2008, 2012, and 2014

Year	Sample Size
2002	11391
2004	8780
2006	8655
2008	9805
2010	8988
2012	9068
2014	8152
2016	7133

Table 3. Summary statistics on key E-FEM populations.

Characteristics	Stock	Replenishing	Transition
Year	2012	2012	2004 - 2016
Wave	6	6	2 - 8
Age, Mean (SD)	66.2 (10.8)	51.5 (0.5)	66.1 (10.4)
Female %	52.9	49.7	53.0
BMI, Mean (SD)	28.4 (5.4)	27.8 (5.4)	28.3 (5.3)
BMI Category (%)
< 25	28.3	37.1	28.0
≥ 25 & < 30	38.8	34.8	39.9
≥ 30	33.0	28.1	32.1
Smoking Status %
Ever Smoked	62.3	44.5	62.3
Current Smoker	13.6	13.7	14.3
Education Level %
Less than Secondary	29.7	11.1	33.1
Upper Secondary and Vocational	59.9	64.1	65.7
Tertiary	10.4	24.8	11.2
Disease Prevalence % (Incidence)
Cancer	9.8	2.9	9.0 (1.46)
Diabetes	11.1	2.8	10.4 (1.34)
Heart Disease	18.9	7.7	18.7 (2.57)
Stroke	4.7	0.0	4.7 (0.85)
Lung Disease	6.1	0.3	6.2 (0.87)
Hypertension	41.8	20.4	42.0 (3.08)
Alzheimer’s	0.35	0.0	0.32 (0.21)
Dementia	1.25	0.27	1.12 (0.58)
High Cholesterol	37.0	14.9	32.5 (3.92)

Table 4. Information on transitioned variables in the model.

Outcome	Variable Type	Regression Model	Predictors*
Outcome	Variable Type	Regression Model	Health Status	Risk Behaviours	Economic Predictors	Demographics
Mortality incidence	Binary Absorbing	Probit	Cancer Diabetes Heart Disease Lung Disease Stroke Dementia	Smoking		Age Sex Education
Cancer incidence	Binary Absorbing	Probit		BMI Smoking Alcohol Consumption		Age Sex Ethnicity Education
Diabetes incidence	Binary Absorbing	Probit	Hypertension High Cholesterol	BMI Physical Activity Alcohol Consumption		Age Sex Ethnicity Education
Heart Disease incidence	Binary Absorbing	Probit	Diabetes Hypertension High Cholesterol	BMI Smoking Physical Activity Alcohol Consumption		Age Sex Ethnicity Education
Hypertension incidence	Binary Absorbing	Probit	High Cholesterol	BMI Smoking Alcohol Consumption Physical Activity		Age Sex Ethnicity Education
Lung Disease incidence	Binary Absorbing	Probit		BMI Smoking		Age Sex Ethnicity Education
Stroke incidence	Binary Absorbing	Probit	Hypertension Diabetes High Cholesterol	BMI Smoking Alcohol Consumption		Age Sex Ethnicity Education
High Cholesterol incidence	Binary Absorbing	Probit		BMI Smoking Physical Activity		Age Sex Ethnicity Education
Dementia incidence	Binary Absorbing	Probit	Hypertension Stroke	BMI Smoking		Age Sex Ethnicity Education
Alzheimers incidence	Binary Absorbing	Probit	Hypertension Stroke	BMI Smoking Alcohol Consumption		Age Sex Ethnicity Education
Start/Stop Smoking	Binary	Probit		BMI		Age Sex Ethnicity Education
Alcohol Consumption	Binary	Probit		BMI Physical Activity		Age Sex Ethnicity Education
BMI	Continuous	OLS		BMI Physical Activity		Age Sex Ethnicity Education
Functional Limitations	Ordered	Oprobit	Stroke Dementia Alzheimers	BMI Smoking		Age Sex Ethnicity Education
Physical Activity	Ordered	Oprobit		Functional Limitations Physical Activity		Age Sex Ethnicity Education
Labour Force Participation	Unordered	Mlogit	Functional Limitations			Age Sex Ethnicity Education

All predictor variables are 2 year lag.

Table 5. Parameter estimates for probit model of heart disease incidence.

	Heart Disease
Male	0.108***
White	-0.0729
Less than Secondary Education	-0.0642
Higher or further education	0.0194
Lag of age spline, less than 65	0.0103
Lag of age spline, 65 to 74	0.0262***
Lag of age spline, more than 75	0.0182***
Lag of BMI spline, less than 30	0.163
Lag of BMI spline, more than 30	0.218
Lag of ever smoker	0.0259
Lag of current smoker	0.0264
Lag of heavy smoker (>20 cigs/day)	0.0836
Lag of problem drinker (>12 drinks/week)	-0.0318
Lag of doctor ever - Diabetes	0.0520
Lag of doctor ever - Hypertension	0.163***
Lag of doctor ever - High Cholesterol	0.0659*
Lag of activity level - Low	0.174**
Lag of activity level - Moderate	0.0979*
Constant	-3.172***
N	22,333
Pseudo R2	0.0335

Table 6. Parameter estimates for probit models of smoking initiation and cessation.

	Smoking
	Started	Stopped
Male	0.112**	0.0263
White	0.0228	0.257
Less than Secondary education	0.188***	-0.106*
Higher or Further education	-0.154**	0.215**
Lag of age spline, less than 65	-0.00572	0.0188***
Lag of age spline, 65 to 74	-0.0186**	-0.00979
Lag of age spline, more than 75	-0.0289**	0.0171
Lag of BMI spline, less than 30	-0.414*	0.839***
Lag of BMI spline, more than 30	0.25	0.0703
Constant	-0.649	-5.060***
N	41951	6157
Pseudo R2	0.019	0.0129

Table 7. Life year and disability-free life year comparison between baseline and no smoking scenario. This is a treatment on the treated analysis, so only including respondents that smoke.

Scenario	Life Years	Disability-free Life Years
Baseline	24.4	16.9
Intervention	31.4	21.4

Table 8. Average improvement in life years and disability-free life years between scenarios by highest level of education.

Education Level	Life Years Improvement	Disability-free Life Years Improvement
1	5.5	3.4
2	8.4	5.3
3	7.7	5.4

Table A1. Summary statistics for raw and imputed stock populations, including the percentage of missing data where applicable.

Characteristics	Raw Stock	Imputed	Missing (%)
Age, Mean (SD)	66.5 (10.8)	66.2 (10.8)	0
Female (%)	52.9	52.9	0
BMI, Mean (SD)	28.4 (5.4)	28.4 (5.4)	19.2
BMI Category (%)
< 25	28.3	28.3	-
≥ 25 & < 30	42.2	42.2	-
≥ 30	29.5	29.5	-
Smoking Status (%)
Ever Smoked	62.5	62.3	1.83
Current Smoker	12.2	13.6	0.05
Education Level (%)			20.9
Less than Secondary	38.6	38.6	-
Upper Secondary and Vocational	48.0	47.9	-
Some University or more	13.5	13.5	-
Initial Disease Prevalence (%)
Cancer	9.8	9.8	0.009
Diabetes	11.1	11.1	0.009
Heart Disease	18.9	18.9	0.009

Table A2. Raw vs imputed BMI mean and standard deviation at each wave of ELSA.

Wave	Raw BMI	Imputed BMI
1	-	28.0 (5.3)
2	28.0 (4.9)	28.0 (5.0)
3	-	28.2 (5.2)
4	28.3 (5.4)	28.4 (5.4)
5	-	28.4 (5.4)
6	28.4 (5.3)	28.4 (5.4)
7	-	28.4 (5.4)
8	28.3 (5.5)	28.3 (5.5)

Table A3. T-test results for health outcomes.

Variable	FEM 3	ELSA 3	P value 3	FEM 5	ELSA 5	P value 5	FEM 8	ELSA 8	P value 8
Alzheimers	0.0067	0.00325	0.00076	0.0103	0.00404	0	0.0178	0.0119	0.0204
Any ADLs	0.184	0.213	0	0.199	0.234	0	0.241	0.225	0.0938
Any IADLs	0.204	0.222	0.012	0.230	0.255	0.00292	0.268	0.266	0.839
Cancer	0.1	0.0863	0.00448	0.130	0.119	0.0836	0.176	0.164	0.182
Dementia	0.0159	0.0102	0.00156	0.0226	0.0194	0.24	0.0324	0.164	0.0214
Diabetes	0.109	0.0937	0.00275	0.132	0.128	0.571	0.164	0.151	0.125
Heart disease	0.220	0.175	0	0.26	0.222	0	0.328	0.29	0.00031
Hypertension	0.453	0.448	0.555	0.491	0.506	0.115	0.549	0.538	0.381
Lung disease	0.0836	0.0719	0.00989	0.0957	0.0863	0.0857	0.111	0.0898	0.00167
Stroke	0.0558	0.0504	0.156	0.0667	0.06743	0.875	0.082	0.089	0.289

Table A3. t-test results for economic outcomes.

Variable	FEM 3	ELSA 3	P value 3	FEM 5	ELSA 5	P value 5	FEM 8	ELSA 8	P value 8
Employed	0.284	0.285	0.906	0.194	0.179	0.0412	0.0809	0.0842	0.606
Unemployed	0.00102	0.00662	0.0132	0.00519	0.00675	0.321	0.00137	0.00092	0.529
Retired/Disabled	0.706	0.707	0.0.84	0.801	0.811	0.167	0.918	0.912	0.411

Table A3. t-test results for risk behaviours.

Variable	FEM 3	ELSA 3	P value 3	FEM 5	ELSA 5	P value 5	FEM 8	ELSA 8	P value 8
BMI*							28.4	28.0	0.0014
Drinks Alcohol	0.866	0.874	0.205	0.859	0.851	0.239	0.844	0.839	0.552
Problem Drinker	0.12	0	0	0.122	0.127	0.487	0.109	0.113	0.567
Activity - Low	0.0907	0.101	0.0426	0.099	0.13	0	0.112	0.129	0.0304
Activity - Moderate	0.164	0.162	0.739	0.168	0.181	0.0775	0.183	0.201	0.0508
Activity - High	0.745	0.736	0.231	0.733	0.689	0	0.706	0.669	0.00097
Smoke Now	0.132	0.143	0.0678	0.106	0.116	0.106	0.076	0.0708	0.391
Smoke Ever	0.637	0.628	0.279	0.639	0.648	0.322	0.638	0.658	0.0646
Heavy Smoker	0.0234	0.0297	0.906	0.0214	0.0229	0.595	0.0128	0.0136	0.761

ELSA only reports BMI information for even waves.

Table A3. t-test results for mortality.

Variable	FEM 3	ELSA 3	P value 3	FEM 5	ELSA 5	P value 5	FEM 8	ELSA 8*	P value 8
Died	0.0294	0.0412	0.00033	0.0435	0.0602	0	0.0596	0	0

End of life interviews are only reporte for waves 2-6 (excluding wave 5). There is therefore no mortality informationfor waves 7 or 8.