Is the task of segmenting a heterogeneous population into a number of more homogeneous subgroups?

  • PDFView PDF

Is the task of segmenting a heterogeneous population into a number of more homogeneous subgroups?

Is the task of segmenting a heterogeneous population into a number of more homogeneous subgroups?

Under a Creative Commons license

Open access

Abstract

Designing policies to harness the potential of heterogeneous target groups such as nonindustrial private forest owners to contribute to public policy goals can be challenging. The behaviors of such groups are shaped by their diverse motivations and circumstances. Segmenting heterogeneous target groups into more homogeneous subgroups may improve the chances of successfully identifying policy strategies to influence their behavior. Findings from a multimethod study of nonindustrial private forest owners in eastern Oregon suggest four unique subgroups of owners with different fuel management motivations and suitabilities for policy tools: commodity managers could benefit from market-based incentives; amenity managers could benefit from capacity building programs paired with symbolic campaigns; recreational managers could benefit from public incentives provided through consultants or contractors who can help plan the work; and passive managers may benefit from opportunities to respond to the policy strategies designed for the other groupings until more information can be gathered. Incorporating qualitative analysis of interview data with statistical analysis of survey data improved understanding of the groupings and appropriate policy strategies for them.

Highlights

► This multi-method study identified four NIPF owner policy target groups ► The policy target groups were commodity, amenity, recreational and passive managers ► Qualitative interview data complimented quantitative survey data in the analysis

Keywords

Policy design

Policy target groups

Segmentation

Nonindustrial private forest owners

Fuel reduction policy

Cited by (0)

Published by Elsevier B.V.

  • Journal List
  • JAMA Network
  • PMC6735407

JAMA Netw Open. 2019 Sep; 2(9): e1910878.

Key Points

Question

What are the sociodemographic characteristics, health conditions, health care utilization, and health care costs for different population segments of a multiethnic Asian population in Singapore divided according to a modified British Columbia Population Segmentation Framework?

Findings

In this cross-sectional study, the study population of 1 181 024 individuals was divided into 8 segments with distinct and heterogeneous sociodemographic characteristics, health conditions, health care utilization, and health care cost patterns.

Meaning

This critical population-level health information can be used as baseline data to inform regional and national health priorities for health services research and policy.

Abstract

Importance

Descriptive population-level health data are critical components of the evidence base for population health policy. Human populations often display marked heterogeneity in their health status among subgroups of the population. The recent widespread adoption of electronic health records provides opportunities to use routine real-world health care data to examine population health.

Objective

To report population sociodemographic characteristics, health conditions, health care utilization, and health care costs for different population segments of a multiethnic Asian population divided according to a modified British Columbia Population Segmentation Framework.

Design, Setting, and Participants

This population-based cross-sectional study used 2016 data from the Singapore Eastern Regional Health System, the largest Regional Health System in Singapore. Data were obtained from deidentified national-level electronic health records at the Ministry of Health Singapore. Participants included all residents in the Singapore Eastern Regional Health System catchment area in 2016. The descriptive analysis was conducted in August 2018.

Main Outcomes and Measures

Socioeconomic profiles, disease prevalence, health care utilization, and cost patterns of population segments.

Results

The total size of the study population in 2016 was 1 181 024 residents (576 663 [48.83%] male; median [interquartile range] age, 40 [22-57] years). The population was divided into 8 segments: healthy with no outpatient utilization (493 483 residents), healthy with outpatient utilization (259 909 residents), healthy with inpatient admissions (49 588 residents), low complex (215 134 residents), medium complex (79 350 residents), high complex (44 445 residents), cancer (34 217 residents), and end of life (4898 residents). Overall, the 3 most prevalent conditions were chronic kidney disease (31.89%), hypertension (18.52%), and lipid disorders (18.33%). Distributions of chronic conditions differed across the segments. Different segments had varying health care utilization patterns: the high-complex segment had the highest number of primary care clinic visits (mean [SD], 4.25 [5.46] visits), the cancer segment had the highest number of specialist outpatient clinic visits (mean [SD], 5.55 [8.49] visits), and the end-of-life segment had the highest numbers of accident and emergency department visits (mean [SD], 1.80 [1.88] visits) and inpatient admissions (mean [SD], 1.99 [1.89] admissions) during 2016. For health care costs, specialist outpatient clinic and inpatient care together made up more than 85% of the total cost of nearly 2 billion Singapore dollars. The end-of-life segment contributed approximately 50% of the health care cost per capita of 60 000 Singapore dollars.

Conclusions and Relevance

Different population segments displayed heterogeneity in sociodemographic characteristics, health conditions, health care utilization, and health care cost patterns. This critical health information can be used as baseline data to inform regional and national health priorities for health services research and policy.

Introduction

Population health is an emerging concept that is defined as “the health outcomes of a group of individuals, including the distribution of such outcomes within the group.”1 Valid and reliable studies2,3,4 of the health states of populations are critical components of the evidence base for population health policy formulation, evaluation of health systems, and development of intervention programs. Interest in population health has been increasing in the medical literature, including many recent population-level studies.5,6,7,8 Population-based approaches to characterizing the complexity of people’s health, including the distribution, variations, and trends of diseases, will make a significant contribution to the field of population health.

Human populations often display marked heterogeneity in their health and disease status, not only between populations but also among subgroups of the same population.9,10 For example, within King County, Washington, there is an approximately 18-year difference in male life expectancy among the US Census tracts, even though the overall health in King County is among the best in the country, ranking in the 95th percentile among county health nationwide.11 Population health science has long focused on aggregate health indicators; however, the study of variability and heterogeneity can yield a deeper understanding of population health characteristics and needs.12 Although such research is challenging, it is important to identify the different health care needs and health characteristics of subgroups within a particular population.13

One promising approach to address the issue of population health heterogeneity is population segmentation. By dividing a heterogeneous population into somewhat homogeneous subgroups, population segmentation helps in understanding the distinctive needs of different parts of the population and developing care interventions and services that different subgroups of people might need.14 Since its conception by Lynn et al,15 population segmentation has been widely applied in many health systems, including British Columbia, Canada,13 Northwest London, United Kingdom,16 and Lombardy, Italy.17 However, data on population segmentation in Asian settings are sparse.

Traditionally, population health studies rely heavily on population-level surveys, such as the National Health Interview Survey by the US Census Bureau18 and National Population Health Survey by Statistics Canada.19 However, population-level surveys are labor intense and time-consuming and rely heavily on self-reported health measures. The interpretation and comparability of these measures are problematic when different persons understand and respond to a given question in different ways.20 The recent widespread adoption of electronic health records (EHRs) globally offers opportunities to use clinical data to examine population health because they provide real-world data available for analysis at any time. The use of EHRs not only facilitates clinical decision-making but also reduces the potential for inaccuracy and incompleteness of data entry and storage.21 In this population-level health study, we report population sociodemographic characteristics, health conditions, health care utilization, and health care costs of different population segments of a multiethnic Asian population divided according to a modified British Columbia Population Segmentation Framework.13

Methods

Data Source

This population-based cross-sectional descriptive study used routinely collected retrospective data from the Ministry of Health Singapore Division of Policy Research and Evaluation. We extracted the following data for each resident in the study population and presented population-level data in this study: demographic characteristics (age, sex, and race/ethnicity), disease diagnosis (as defined by diagnosis codes from the International Classification of Diseases, Tenth Revision), and health care utilization nationwide, including polyclinic visits, general practitioner visits, specialist outpatient clinic visits, accident and emergency department visits, and inpatient admissions. Polyclinics are large public primary care clinics located throughout the country that provide subsidized primary care, including medical treatment, preventive health care, and health education.22 Charlson Comorbidity Index23 and Elixhauser Comorbidity Index24 scores were also calculated for overall population and each population segment. The data sets captured comprehensive data and covered health facility visits to all health care professionals and to public health care institutions, all hospitalizations and outpatient operations in private hospitals, and outpatient visits that are covered under MediSave (Singapore’s national medical savings account program) and MediShield (Singapore’s national health insurance program). Data from general practitioners were included if they participate in Community Health Assist Scheme, a government program that enables Singapore citizens from low- and middle-income households and those born on or before December 31, 1949, to receive subsidies for medical and dental care.

This study was exempted from formal review by the SingHealth Centralized Institutional Review Board. The need for informed consent was waived because the data were deidentified. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies.25

Study Population

The study population comprised all residents who lived in the Singapore Eastern Regional Health System (RHS) in 2016. The Eastern RHS is the largest RHS established by the Ministry of Health Singapore to integrate health service across health care systems in the Eastern region across care settings—primary care, intermediate care, long-term care, community-based care, and tertiary care—to deliver comprehensive care to the regional population in eastern Singapore.26 The catchment population as of June 2017 was approximately 1 336 840.27

Population Segmentation

The population segmentation approach used in this study was based on the British Columbia Population Segmentation Framework.13 The framework was chosen and modified by a panel of local experts in Singapore who are senior clinicians and health policy leaders in the Eastern RHS and Ministry of Health Singapore. They come from a variety of specialties with extensive experience in health policy and clinical care. This ensures that policy implementation is relevant to the specific health care system setting within Singapore. The definitions and criteria for the modified British Columbia Population Segmentation Framework are finalized via focused group discussions among the Ministry of Health Singapore and Eastern RHS experts (Table 1).

Table 1.

Modified British Columbia Population Segmentation Framework

CategoryDescription
End of life Patients who died in the last 1 y and had received a diagnosis of high complex diseases and/or cancer
Cancer Cancer or metastatic cancer
High complex Dementia, heart failure, kidney transplant, severe liver disease OR had this event or intervention (stroke or dialysis) OR had this combination of conditions (angina and chronic obstructive pulmonary disease; acute myocardial infarction and predialysis chronic kidney disease; rheumatoid arthritis and osteoporosis; diabetes, hypertension, and osteoarthritis)
Medium complex Predialysis chronic kidney disease, chronic obstructive pulmonary disease, coronary heart disease including angina, rheumatoid arthritis, Parkinson disease, atrial fibrillation, moderate liver disease, peripheral vascular disease, OR had this event or intervention (coronary artery bypass grafting, acute myocardial infarction, percutaneous coronary intervention, and lower extremity amputation) OR had this combination of conditions (osteoporosis and osteoarthritis, osteoporosis and hypertension, osteoarthritis and hypertension, osteoporosis and spine or hip fracture)
Low complex Diabetes, hypertension, osteoporosis, osteoarthritis, asthma, epilepsy, hyperthyroidism, hypothyroidism, benign prostatic hypertrophy, and lipid disorders
Healthy with inpatient admissions Not end of life, cancer, high, medium, or low complex but with ≥1 inpatient admission
Healthy with outpatient utilization Not end of life, cancer, high, medium, or low complex but with ≥1 polyclinic or specialist outpatient clinic visit and no inpatient admissions
Healthy with no outpatient utilization Not end of life, cancer, high, medium, or low complex with no polyclinic visit, specialist outpatient clinic visit, and inpatient admissions

Statistical Analysis

The descriptive analysis was conducted in August 2018. Means, SDs, and percentages were used as appropriate to describe the data. Stata statistical software version 14.0 (StataCorp) was used for statistical analysis, and Tableau statistical software version 2018.2.2 (Tableau Software) was used for data visualization.

Results

The total size of the study population in 2016 was 1 181 024 residents. As shown in Table 2, the population was divided into 8 segments—healthy with no outpatient utilization (493 483 residents), healthy with outpatient utilization (259 909 residents), healthy with inpatient admissions (49 588 residents), low complex (215 134 residents), medium complex (79 350 residents), high complex (44 445 residents), cancer (34 217 residents), and end of life (4898 residents)—on the basis of the modified British Columbia Population Segmentation Framework.13 The largest segment was healthy with no outpatient utilization (493 483 residents) and the smallest was end of life (4898 residents).

Table 2.

Population Health Characteristics in 2016, by Population Segment

CharacteristicTotalHealthy With No Outpatient UtilizationHealthy With Outpatient UtilizationHealthy With Inpatient AdmissionsLow ComplexMedium ComplexHigh ComplexCancerEnd of Life
Population size, No. 1 181 024 493 483 259 909 49 588 215 134 79 350 44 445 34 217 4898
Sociodemographic characteristics
Male, No. (%) 576 663 (48.83) 249 375 (50.53) 123 180 (47.39) 19 964 (40.26) 102 232 (47.52) 44 908 (56.59) 21 746 (48.93) 12 743 (37.24) 2515 (51.35)
Age, median (IQR), y 40 (22-57) 33 (18-46) 28 (15-45) 29 (6-38) 55 (41-64) 63 (54-72) 69 (60-78) 63 (54-73) 78 (67-86)
Residential status, No. (%)
Permanent resident 106 034 (8.98) 64 952 (13.16) 18 982 (7.30) 6447 (13.00) 9440 (4.39) 3268 (4.12) 1073 (2.41) 1746 (5.10) 126 (2.57)
Singapore citizen 1 060 860 (89.83) 420 324 (85.17) 236 848 (91.13) 42 131 (84.96) 205 272 (95.42) 75 952 (95.72) 43 228 (97.26) 32 395 (94.68) 4710 (96.16)
Ethnic group, No. (%)
Chinese 827 811 (70.09) 338 184 (68.53) 179 704 (69.14) 33 520 (67.60) 154 052 (71.61) 58 463 (73.68) 31 881 (71.73) 28 318 (82.76) 3689 (75.32)
Indian 84 104 (7.12) 33 539 (6.80) 18 560 (7.14) 3679 (7.42) 16 154 (7.51) 6361 (8.02) 3880 (8.73) 1648 (4.82) 283 (5.78)
Malay 158 907 (13.46) 56 866 (11.52) 39 914 (15.36) 8062 (16.26) 32 636 (15.17) 10 813 (13.63) 7161 (16.11) 2903 (8.48) 552 (11.27)
Others 109 771 (9.29) 64 893 (13.15) 21 731 (8.36) 4273 (8.62) 12 214 (5.68) 3614 (4.55) 1391 (3.13) 1281 (3.74) 374 (7.64)
Community Health Assist Scheme status of not enrolled, No. (%)a 777 007 (65.79) 364 384 (73.84) 176 021 (67.72) 39 196 (79.04) 115 830 (53.84) 41 091 (51.78) 19 060 (42.88) 19 104 (55.83) 2321 (47.39)
Prevalence of diseases or conditions, No. (%)
Angina 22 554 (1.91) 0 0 0 0 13 802 (17.39) 6354 (14.30) 1763 (5.15) 635 (12.96)
Anxiety 15 687 (1.33) 2317 (0.47) 3320 (1.28) 700 (1.41) 4822 (2.24) 2233 (2.81) 1309 (2.95) 845 (2.47) 141 (2.88)
Arrhythmia 18 371 (1.56) 1588 (0.32) 1485 (0.57) 410 (0.83) 2861 (1.33) 4804 (6.05) 4628 (10.41) 1519 (4.44) 1 076 (21.97)
Asthma 60 634 (5.13) 0 0 0 46 790 (21.75) 7072 (8.91) 4276 (9.62) 2078 (6.07) 418 (8.53)
Atrial fibrillation 14 316 (1.21) 0 0 0 0 6027 (7.60) 5675 (12.77) 1371 (4.01) 1243 (25.38)
Benign prostatic hyperplasia 22 322 (1.89) 0 0 0 8326 (3.87) 5803 (7.31) 4180 (9.40) 3236 (9.46) 777 (15.86)
Chronic obstructive pulmonary disease 10 447 (0.88) 0 0 0 0 6328 (7.97) 2560 (5.76) 983 (2.87) 576 (11.76)
Chronic heart disease 58 589 (4.96) 0 0 0 0 34 089 (42.96) 17 231 (38.77) 4832 (14.12) 2437 (49.76)
Diabetes 102 738 (8.70) 0 0 0 46 543 (21.63) 18 735 (23.61) 26 893 (60.51) 8222 (24.03) 2345 (47.88)
Lipid disorders 216 510 (18.33) 0 0 0 109 412 (50.86) 51 395 (64.77) 36 187 (81.42) 16 253 (47.50) 3263 (66.62)
Epilepsy 8429 (0.71) 0 0 0 5669 (2.64) 692 (0.87) 1354 (3.05) 459 (1.34) 255 (5.21)
Gestational diabetes 7633 (0.65) 1980 (0.40) 1500 (0.58) 957 (1.93) 2549 (1.18) 313 (0.39) 172 (0.39) 157 (0.46) 5 (0.10)
Hemorrhagic stroke 4978 (0.42) 0 0 0 0 0 4166 (9.37) 376 (1.10) 436 (8.90)
Heart failure 13 889 (1.18) 0 0 0 0 0 11 033 (24.82) 1281 (3.74) 1575 (32.16)
Hip fracture 6870 (0.58) 374 (0.08) 325 (0.13) 119 (0.24) 1158 (0.54) 1455 (1.83) 2174 (4.89) 680 (1.99) 585 (11.94)
Hypertension 218 767 (18.52) 0 0 0 102 831 (47.80) 55 927 (70.48) 39 044 (87.85) 17 079 (49.91) 3886 (79.34)
Hyperthyroidism 12 239 (1.04) 0 0 0 7990 (3.71) 1800 (2.27) 1396 (3.14) 883 (2.58) 170 (3.47)
Hypothyroidism 12 329 (1.04) 0 0 0 6460 (3.00) 2120 (2.67) 1859 (4.18) 1576 (4.61) 314 (6.41)
Insomnia 11 800 (1.00) 1405 (0.28) 2731 (1.05) 417 (0.84) 3599 (1.67) 1707 (2.15) 1136 (2.56) 655 (1.91) 150 (3.06)
Ischemic stroke 14 527 (1.23) 0 0 0 0 0 12 190 (27.43) 1254 (3.66) 1083 (22.11)
Myocardial infarction 19 312 (1.64) 0 0 0 0 8900 (11.22) 7518 (16.92) 1353 (3.95) 1541 (31.46)
Nephritis 5925 (0.50) 7 (0.00)b 41 (0.02) 21 (0.04) 311 (0.14) 3054 (3.85) 1784 (4.01) 526 (1.54) 181 (3.70)
Nephrosis 3145 (0.27) 9 (0.00)b 66 (0.03) 26 (0.05) 310 (0.14) 965 (1.22) 1286 (2.89) 301 (0.88) 182 (3.72)
Osteoarthritis 66 968 (5.67) 0 0 0 19 909 (9.25) 22 092 (27.84) 18 799 (42.30) 5128 (14.99) 1040 (21.23)
Osteoporosis 1811 (0.15) 0 0 0 127 (0.06) 641 (0.81) 660 (1.48) 222 (0.65) 161 (3.29)
Parkinson disease 3128 (0.26) 0 0 0 0 1599 (2.02) 983 (2.21) 286 (0.84) 260 (5.31)
Peripheral vascular disease 5196 (0.44) 0 0 0 0 2163 (2.73) 2062 (4.64) 439 (1.28) 532 (10.86)
Prediabetes 10 862 (0.92) 298 (0.06) 229 (0.09) 58 (0.12) 4031 (1.87) 2571 (3.24) 2479 (5.58) 937 (2.74) 259 (5.29)
Hypertension during pregnancy 1958 (0.17) 476 (0.10) 299 (0.12) 337 (0.68) 714 (0.33) 78 (0.10) 34 (0.08) 19 (0.06) 1 (0.02)
Psoriasis 2602 (0.22) 261 (0.05) 490 (0.19) 77 (0.16) 752 (0.35) 485 (0.61) 354 (0.80) 139 (0.41) 44 (0.90)
Renal disease 36 539 (3.09) 0 0 0 0 19 194 (24.19) 11 885 (26.74) 3437 (10.04) 2023 (41.30)
Respiratory failure 4870 (0.41) 104 (0.02) 93 (0.04) 56 (0.11) 442 (0.21) 859 (1.08) 1946 (4.38) 524 (1.53) 846 (17.27)
Rheumatoid arthritis 2895 (0.25) 0 0 0 0 2041 (2.57) 541 (1.22) 250 (0.73) 63 (1.29)
Secondary hypertension 410 (0.03) 8 (0.00)b 16 (0.01) 13 (0.03) 56 (0.03) 87 (0.11) 167 (0.38) 45 (0.13) 18 (0.37)
Spine fracture 4900 (0.41) 749 (0.15) 499 (0.19) 158 (0.32) 1056 (0.49) 846 (1.07) 990 (2.23) 404 (1.18) 198 (4.04)
Stroke 18 120 (1.53) 0 0 0 0 0 15 254 (34.32) 1527 (4.46) 1339 (27.34)
Weight loss 14 718 (1.25) 1023 (0.21) 1078 (0.41) 354 (0.71) 5787 (2.69) 2828 (3.56) 2733 (6.15) 765 (2.24) 150 (3.06)
Bipolar disorder 1762 (0.15) 215 (0.04) 357 (0.14) 131 (0.26) 555 (0.26) 200 (0.25) 195 (0.44) 90 (0.26) 19 (0.39)
Dementia 6977 (0.59) 0 0 0 0 0 5209 (11.72) 709 (2.07) 1059 (21.62)
General anxiety disease 1469 (0.12) 88 (0.02) 376 (0.14) 63 (0.13) 470 (0.22) 237 (0.30) 156 (0.35) 69 (0.20) 10 (0.20)
Major depression 21 655 (1.83) 3301 (0.67) 4139 (1.59) 1149 (2.32) 5789 (2.69) 2674 (3.37) 2901 (6.53) 1194 (3.49) 508 (10.37)
Schizophrenia 4942 (0.42) 516 (0.10) 1011 (0.39) 218 (0.44) 1736 (0.81) 640 (0.81) 535 (1.20) 223 (0.65) 63 (1.29)
Cancer without metastasis 36 022 (3.05) 0 0 0 0 0 0 33 647 (98.33) 2375 (48.49)
Metastatic carcinoma 10 382 (0.88) 0 0 0 0 0 0 8822 (25.78) 1560 (31.85)
Moderate liver disease 2514 (0.21) 0 0 0 0 770 (0.97) 889 (2.00) 541 (1.58) 314 (6.41)
Severe liver disease 1422 (0.12) 0 0 0 0 0 905 (2.04) 288 (0.84) 229 (4.68)
Coronary artery bypass graft 4760 (0.40) 0 0 0 0 2509 (3.16) 1759 (3.96) 338 (0.99) 154 (3.14)
Percutaneous coronary intervention 17 206 (1.46) 0 0 0 0 10 840 (13.66) 4789 (10.78) 1139 (3.33) 438 (8.94)
Chronic kidney disease, receiving dialysis or predialysis 3467 (0.29) 0 0 0 0 0 2534 (5.70) 432 (1.26) 501 (10.23)
Kidney transplant 99 (0.01) 0 0 0 0 0 95 (0.21) 3 (0.01) 1 (0.02)
Major lower extremity amputation 714 (0.06) 0 0 0 0 268 (0.34) 269 (0.61) 51 (0.15) 126 (2.57)
Minor lower extremity amputation 1341 (0.11) 0 0 0 0 575 (0.72) 549 (1.24) 76 (0.22) 141 (2.88)
Chronic kidney disease 376 590 (31.89) 48 451 (9.82) 58 063 (22.34) 18 267 (36.84) 121 512 (56.48) 61 020 (76.90) 39 456 (88.77) 25 331 (74.03) 4490 (91.67)
Charlson Comorbidity Index score, mean (SD) 0.65 (1.37) 0.10 (0.30) 0.22 (0.01) 0.37 (0.02) 0.78 (0.15) 1.76 (1.79) 3.08 (4.18) 4.95 (8.27) 6.75 (12.13)
Elixhauser Comorbidity Index score, mean (SD) 0.60 (2.56) 0.01 (0.46) 0.42 (0.65) 0.49 (0.82) 0.71 (1.22) 1.19 (2.91) 1.65 (5.24) 3.07 (6.19) 3.46 (8.13)
Health care facility visits, mean (SD), No.
Polyclinic visits 1.51 (3.27) 0 1.87 (2.81) 1.60 (4.03) 2.91 (3.97) 3.73 (4.94) 4.25 (5.46) 2.88 (4.36) 1.27 (3.58)
Specialist outpatient clinic visits 1.11 (2.84) 0 1.23 (1.90) 2.21 (3.50) 1.42 (2.59) 2.54 (3.58) 3.79 (4.53) 5.55 (8.49) 4.10 (7.24)
Accident and emergency department visits 0.23 (0.82) 0.06 (0.30) 0.25 (0.75) 0.80 (1.38) 0.22 (0.78) 0.37 (1.04) 0.74 (1.98) 0.38 (1.07) 1.80 (1.88)
Inpatient admissions 0.13 (0.55) 0 0 1.23 (0.70) 0.11 (0.45) 0.26 (0.72) 0.61 (1.27) 0.46 (1.19) 1.99 (1.89)

Sociodemographic Characteristics

The median (interquartile range) age of the study population was 40 (22-57) years, and 576 663 participants (48.83%) were male. Table 2 shows that the segment with cancer had a disproportionately small percentage of male participants (37.24%). In the total population, most (70.09%) were ethnically Chinese. The proportion of Chinese participants was higher in the low-complex, medium-complex, high-complex, cancer, and end-of-life segments. These segments also had older participants and a higher proportion of residents enrolled in Community Health Assist Scheme compared with the overall population.

The Figure shows the proportion of patients in different age groups within each segment. Healthier segments had high proportions of patients younger than 44 years, whereas the complex, cancer, and end-of-life segments had high proportions of patients between ages 45 and 84 years. The end-of-life segment had the highest proportion of patients aged 65 years and older.

Is the task of segmenting a heterogeneous population into a number of more homogeneous subgroups?

Proportion of Patients in Different Age Groups Within Each Segment

Graphs show proportion of patients in different age groups within each segment. Healthier segments have high proportions of patients younger than 44 years, whereas the complex, cancer, and end-of-life segments have high proportions of patients ages 45 to 84 years. The end-of-life segment has the highest proportion of patients age 65 years and older.

Health Conditions

Overall, the 5 most prevalent diseases were chronic kidney disease (31.89%), hypertension (18.52%), lipid disorders (18.33%), diabetes (8.70%), and osteoarthritis (5.67%). In Table 2, for the first 3 segments (healthy with no outpatient utilization, healthy with outpatient utilization, and healthy with inpatient admissions), the prevalence of chronic diseases was generally lower than that in the overall population. The other 5 segments (low complex, medium complex, high complex, cancer, and end of life) had higher chronic disease prevalence compared with the population mean. For most of the diseases, prevalence increased as population segments became more complex, from the healthy with no outpatient utilization segment to the end-of-life segment. Similar trends were observed with Charlson Comorbidity Index and Elixhauser Comorbidity Index scores.

The distribution of chronic conditions across the different segments is shown in Table 3. As expected, most diseases concentrated in specific segments according to the modified British Columbia Population Segmentation Framework.13 For example, cancer without metastasis was defined under the cancer category in the modified British Columbia Population Segmentation Framework, as shown in Table 1, and 93.41% of patients with cancer without metastasis were classified into the cancer segment. Some diseases were distributed more widely into multiple segments. For example, although osteoarthritis was defined under the low-complex category, 32.99% of patients with osteoarthritis were in the medium-complex segment and another 28.07% were in the high-complex segment.

Table 3.

Distribution of Chronic Conditions Listed in the Modified British Columbia Population Segmentation Framework, by Population Segment

ConditionPatients, %aPatients, No.
Medium ComplexHigh ComplexCancerEnd of Life
Minimum criteria for low complex chronic conditions
Diabetes 18.24 26.18 8.00 2.28 102 738
Hypertension 25.56 17.85 7.81 1.78 218 767
Osteoporosis 35.39 36.44 12.26 8.89 1811
Osteoarthritis 32.99 28.07 7.66 1.55 66 968
Asthma 11.66 7.05 3.43 0.69 60 634
Epilepsy 8.21 16.06 5.45 3.03 8429
Hyperthyroidism 14.71 11.41 7.21 1.39 12 239
Hypothyroidism 17.20 15.08 12.78 2.55 12 329
Benign prostatic hyperplasia 26.00 18.73 14.50 3.48 22 322
Lipid disorders 23.74 16.71 7.51 1.51 216 510
Minimum criteria for medium complex chronic conditions
Chronic kidney disease receiving dialysis or predialysis 0.00 73.09 12.46 14.45 3467
Chronic obstructive pulmonary disease 60.57 24.50 9.41 5.51 10 447
Coronary heart disease 58.18 29.41 8.25 4.16 58 589
Angina 61.20 28.17 7.82 2.82 22 554
Rheumatoid arthritis 70.50 18.69 8.64 2.18 2895
Parkinson disease 51.12 31.43 9.14 8.31 3128
Atrial fibrillation 42.10 39.64 9.58 8.68 14 316
Moderate liver disease 30.63 35.36 21.52 12.49 2514
Peripheral vascular disease 41.63 39.68 8.45 10.24 5196
Coronary artery bypass graft 52.71 36.95 7.10 3.24 4760
Myocardial infarction 46.09 38.93 7.01 7.98 19 312
Percutaneous coronary intervention 63.00 27.83 6.62 2.55 17 206
Major lower extremity amputation 37.54 37.68 7.14 17.65 714
Minor lower extremity amputation 42.88 40.94 5.67 10.51 1341
Minimum criteria for high complex chronic conditions
Dementia 74.66 10.16 15.18 6977
Heart failure 79.44 9.22 11.34 13 889
Kidney transplant 95.96 3.03 1.01 99
Severe liver disease 63.64 20.25 16.10 1422
Stroke 84.18 8.43 7.39 18 120
Minimum criteria for cancer
Cancer without metastasis 93.41 6.59 36 022
Metastatic carcinoma 84.97 15.03 10 382
Total 20.99 11.76 9.05 1.30 378 044

Health Care Utilization

As shown in Table 2, overall, residents in the Eastern Singapore RHS had a mean (SD) of 1.51 (3.27) polyclinic visits, 1.11 (2.84) specialist outpatient clinic visits, 0.23 (0.82) emergency department visit, and 0.13 (0.55) inpatient admission in 2016. The high-complex segment had the highest number of polyclinic visits (mean [SD], 4.25 [5.46] visits). The cancer segment had the highest specialist outpatient clinic visits (mean [SD], 5.55 [8.49] visits). The end-of-life segment had a mean (SD) of 1.80 (1.88) emergency department visits and 1.99 (1.89) inpatient admissions, both of which were the highest among all segments.

Health Care Cost

The distribution of health care cost in 2016 across different segments is shown in Table 4. The healthy with inpatient admission segment accounted for the largest proportion of total cost (21.56%), even though its population size was less than 5% of the total population. Specialist outpatient clinic and inpatient admission costs together accounted for more than 85% of the total cost of nearly 2 billion Singapore dollars.

Table 4.

Distribution of Health Care Cost in 2016, by Population Segment

SegmentPatients, No. (%)Total Polyclinic Visit Cost, SGDaTotal GP Visit Cost, SGDaTotal Specialist Outpatient Clinic Visit Cost, SGDaTotal Accident and Emergency Department Visit Cost, SGDaTotal Inpatient Admission Cost, SGDaTotal Utilization Cost, SGDaTotal Utilization Cost by Segment, %
Healthy with no outpatient utilization 493 483 (41.78) 0 4.82 0 8.55 0 13.37 67
Healthy with outpatient utilization 259 909 (22.01) 31.27 3.77 109.13 19.49 0 163.65 8.18
Healthy with inpatient admissions 49 588 (4.20) 5.89 0.57 36.17 11.73 376.89 431.24 21.56
Low complex 215 134 (18.22) 50.73 19.64 101.39 16.52 149.95 338.22 16.91
Medium complex 79 350 (6.72) 28.60 10.17 65.83 11.09 182.53 298.22 14.91
High complex 44 445 (3.76) 21.97 6.69 55.20 12.92 247.58 344.35 17.22
Cancer 34 217 (2.90) 8.97 3.43 75.55 4.75 176.06 268.76 13.44
End of life 4898 (0.41) 0.54 0.35 8.73 3.53 128.90 142.05 7.10
Total 1 181 024 (100) 147.98 49.44 451.99 88.57 1261.90 1999.87 100.00
Total utilization cost by service, % 7.40 2.47 22.60 4.43 63.10 100.00

The distribution of health care cost per capita in 2016 across different segments is illustrated in eTable 1 in the Supplement. Health care costs per patient in the end-of-life segment were the largest (approximately 50%, or 60 000 Singapore dollars) compared with the other segments.

The distribution of health care cost per capita in 2016 in different age groups across segments is shown in eTable 2 in the Supplement. There seemed to be a general trend that older patients had higher health care cost per capita. Per capita cost for patients in the low-complex segment and healthy with or without outpatient utilization segments were generally low across all age groups. Patients in the end-of-life segment had high health cost per capita across all age groups.

Discussion

In this article, we describe the heterogeneous health profiles of population segments in the Singapore Eastern RHS based on real-world data from routine EHRs. We found that different segments within a population have different sociodemographic characteristics, health conditions, health care utilization, and health care costs. This article provides a critical overview of and baseline information on health status at a population level to inform population health management. For example, compared with the Singapore National Health Survey, which reported a crude prevalence of diabetes of 11.3% among Singapore residents aged 18 to 69 years in 2010,28 our data show that the prevalence of diabetes among residents in the Singapore Eastern RHS was 8.70% in 2016. The lower diabetes prevalence might be multifactorial, including recent national-level concerted health promotion efforts by the Ministry of Health Singapore and different data-collection methods. The data presented here may serve as a starting point for further population health monitoring, care intervention, and policy evaluation. Future research opportunities include tracking the longitudinal health care utilization and cost of these population segments, examining the predictive ability of segment membership on health care utilization, and analyzing how individuals move between the population segments in the long term.

The present study adds to the emerging field of using real-world health data by demonstrating that routine clinical care data can be used for population health profiling and monitoring. Recently, there has been substantial enthusiasm for using routine health care data to provide real-world evidence; however, the widespread use of EHRs, administrative databases, and smart devices has heretofore not led to wide use of these data for clinical insights at the population health level, despite the obvious advantages of using routine care data, which is that the use of such data is both pragmatic and generalizable.29 Real-world data provide useful information that may yield new insights and stimulate new research questions and aid in the design of other interventional studies, such as a clinical trial, and facilitate policy making and practice.29,30 For example, our study found that the healthy with inpatient admission segment unexpectedly accounted for the largest proportion of total cost (21.56%), even though its population size was less than 5% of the total population. This finding warrants further research efforts in understanding the mechanism behind this phenomenon. One explanation for the participants’ health behavior is their private health insurance plan status, which might lead to higher inpatient service rates even though they do not have complex chronic conditions.

The optimal framework for and measures of population health status have long been debated. It is recognized that traditional health indicators alone, such as life expectancy, may not be sufficient for measuring and monitoring population health.31 Functional, psychosocial, quality-of-life, and behavioral variables have been proposed and used as important population health indicators.32 The US Institute of Medicine has also given serious attention to population health measurement and encouraged summary measures that include health-related quality-of-life data.33 Furthermore, a population health perspective also requires attention to resource utilization and allocation.1 Some summary measures have been developed to include a broad spectrum of health indicators in different policy and research contexts, such as EQ-5D34 and the Health Utilities Index.35 Although these data are often made available via population-based health surveys, they may not be routinely collected and stored in many health systems, including our data sets. To further complicate this matter, there is heterogeneity in data definition, collection, and reporting in the current medical literature of population health policy. One of the more widely accepted models is the Organisation for Economic Co-operation and Development health indicators framework,36 which reports health-related data in the following 5 categories: health status, risk factors for health, access to care, quality and outcome of care, and health care resources. However, controversy remains over what the universally important, relevant, and practical health indicators and performance measures are that researchers and policy makers should use, both conceptually and at the implementation level. This is likely dependent on each health care system’s policy priority within their economic and political context, objectives of the funding policy, and data availability.

Limitations

This study has limitations. First, it is retrospective and relies on existing data sources. Second, our study may not be generalizable to other health care system settings because they may not be collecting the exact same type of data that we presented and, thus, may not find our study readily replicable; however, they may find our segmentation framework useful as a reference for their policy needs. Third, our database does not include patients who are exclusively managed in the private health care sector, which was small because MediSave is Singapore’s mandatory national medical savings account program and MediShield is Singapore’s national health insurance program that automatically covers all Singapore residents.

Population health measures need to be developed and validated in close consultation with stakeholders, taking local policy context and priorities into considerations. It remains a critical task for the field of population health research to determine what to measure and how to measure it, whether by population surveys or routine clinical databases. Future research efforts are ongoing to link our routine health care data with collaborators’ databases that are rich in social and behavioral data to further refine population segmentation frameworks.

Conclusions

There are wide differences in the health characteristics of the Eastern RHS population in Singapore. Different population segments displayed heterogeneity in sociodemographic characteristics, health conditions, health care facility utilization, and cost. These data can be used to inform regional and national health priorities for health services research and policy.

Notes

Supplement.

eTable 1. Distribution of Health Cost Per Capita in 2016 Across Different Segments

eTable 2. Distribution of Health Cost Per Capita in 2016 in Different Age Groups, Across Segments

References

2. Murray CJL, Salomon JA, Mathers CD, Lopez AD Summary measures of population health: concepts, ethics, measurement and applications. https://apps.who.int/iris/handle/10665/42439. Published 2002. Accessed June 1, 2018.

3. Evans DB, Tandon A, Murray CJL, Lauer JA. Comparative efficiency of national health systems: cross national econometric analysis. BMJ. 2001;323(7308):307-310. doi: 10.1136/bmj.323.7308.307 [PMC free article] [PubMed] [CrossRef] [Google Scholar]

4. Yan S, Seng BJJ, Kwan YH, et al.. Identifying heterogeneous health profiles of primary care utilizers and their differential healthcare utilization and mortality: a retrospective cohort study. BMC Fam Pract. 2019;20(1):54. doi: 10.1186/s12875-019-0939-2 [PMC free article] [PubMed] [CrossRef] [Google Scholar]

5. Schieber LZ, Guy GP Jr, Seth P, et al.. Trends and patterns of geographic variation in opioid prescribing practices by state, United States, 2006-2017. JAMA Netw Open. 2019;2(3):e190665. doi: 10.1001/jamanetworkopen.2019.0665 [PMC free article] [PubMed] [CrossRef] [Google Scholar]

6. Bilsteen JF, Taylor-Robinson D, Børch K, Strandberg-Larsen K, Nybo Andersen AM. Gestational age and socioeconomic achievements in young adulthood: a Danish population-based study. JAMA Netw Open. 2018;1(8):e186085. doi: 10.1001/jamanetworkopen.2018.6085 [PMC free article] [PubMed] [CrossRef] [Google Scholar]

7. Jensen MV, Rugbjerg K, de Fine Licht S, et al.. Endocrine late effects in survivors of cancer in adolescence and young adulthood: a Danish population-based cohort study. JAMA Netw Open. 2018;1(2):e180349. doi: 10.1001/jamanetworkopen.2018.0349 [PMC free article] [PubMed] [CrossRef] [Google Scholar]

8. Low LL, Yan S, Kwan YH, Tan CS, Thumboo J. Assessing the validity of a data driven segmentation approach: a 4 year longitudinal study of healthcare utilization and mortality. PLoS One. 2018;13(4):e0195243. doi: 10.1371/journal.pone.0195243 [PMC free article] [PubMed] [CrossRef] [Google Scholar]

9. Yan S, Kwan YH, Tan CS, Thumboo J, Low LL. A systematic review of the clinical application of data-driven population segmentation analysis. BMC Med Res Methodol. 2018;18(1):121. doi: 10.1186/s12874-018-0584-9 [PMC free article] [PubMed] [CrossRef] [Google Scholar]

10. Evans R, Barer M, Marmor T. Why Are Some People Healthy and Others Not? The Determinants of Health of Populations. New York, NY: Routledge; 2017. doi: 10.4324/9781315135755 [CrossRef] [Google Scholar]

11. Dwyer-Lindgren L, Stubbs RW, Bertozzi-Villa A, et al.. Variation in life expectancy and mortality by cause among neighbourhoods in King County, WA, USA, 1990-2014: a census tract-level analysis for the Global Burden of Disease Study 2015. Lancet Public Health. 2017;2(9):e400-e410. doi: 10.1016/S2468-2667(17)30165-2 [PubMed] [CrossRef] [Google Scholar]

12. Galea S. Health haves, health have nots, and heterogeneity in population health. Lancet Public Health. 2017;2(9):e388-e389. doi: 10.1016/S2468-2667(17)30160-3 [PubMed] [CrossRef] [Google Scholar]

13. British Columbia Ministry of Health The Health System Matrix 6.1: Understanding the Health Care Needs of the British Columbia Population Through Population Segmentation. Vancouver, BC: British Columbia Ministry of Health; 2015. [Google Scholar]

15. Lynn J, Straube BM, Bell KM, Jencks SF, Kambic RT. Using population segmentation to provide better health care for all: the “Bridges to Health” model. Milbank Q. 2007;85(2):185-208. doi: 10.1111/j.1468-0009.2007.00483.x [PMC free article] [PubMed] [CrossRef] [Google Scholar]

17. Madotto F, Riva MA, Fornari C, et al.. Administrative databases as a tool for identifying healthcare demand and costs in an over-one million population. Epidemiol Biostat Public Health. 2013;10(2):e8840. doi: 10.2427/8840 [CrossRef] [Google Scholar]

18. Blackwell DL, Lucas JW, Clarke TC. Summary health statistics for U.S. adults: National Health Interview Survey, 2012. Vital Health Stat. 2014;260:1-161. [PubMed] [Google Scholar]

19. Dunlop S, Coyte PC, McIsaac W. Socio-economic status and the utilisation of physicians’ services: results from the Canadian National Population Health Survey. Soc Sci Med. 2000;51(1):123-133. doi: 10.1016/S0277-9536(99)00424-4 [PubMed] [CrossRef] [Google Scholar]

20. Mathers CD, Murray CJ, Ezzati M, Gakidou E, Salomon JA, Stein C. Population health metrics: crucial inputs to the development of evidence for health policy. Popul Health Metr. 2003;1(1):6. doi: 10.1186/1478-7954-1-6 [PMC free article] [PubMed] [CrossRef] [Google Scholar]

21. Scholte M, van Dulmen SA, Neeleman-Van der Steen CW, van der Wees PJ, Nijhuis-van der Sanden MW, Braspenning J. Data extraction from electronic health records (EHRs) for quality measurement of the physical therapy process: comparison between EHR data and survey data. BMC Med Inform Decis Mak. 2016;16(1):141. doi: 10.1186/s12911-016-0382-4 [PMC free article] [PubMed] [CrossRef] [Google Scholar]

23. Charlson M, Szatrowski TP, Peterson J, Gold J. Validation of a combined comorbidity index. J Clin Epidemiol. 1994;47(11):1245-1251. doi: 10.1016/0895-4356(94)90129-5 [PubMed] [CrossRef] [Google Scholar]

25. von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP; STROBE Initiative . The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Ann Intern Med. 2007;147(8):573-577. doi: 10.7326/0003-4819-147-8-200710160-00010 [PubMed] [CrossRef] [Google Scholar]

29. Jarow JP, Lavange L, Woodcock J. Multidimensional evidence generation and FDA regulatory decision making: defining and using “real-world” data. JAMA. 2017;318(8):703-704. doi: 10.1001/jama.2017.9991 [PubMed] [CrossRef] [Google Scholar]

30. Schull M, McGrail K, Paprica A. The nascent Pan-Canadian Real-world Health Data Network (PRHDN). Int J Popul Data Sci. 2017;1:125. doi: 10.23889/ijpds.v1i1.144 [CrossRef] [Google Scholar]

31. Macran S, Weatherly H, Kind P. Measuring population health: a comparison of three generic health status measures. Med Care. 2003;41(2):218-231. doi: 10.1097/01.MLR.0000044901.57067.19 [PubMed] [CrossRef] [Google Scholar]

33. Institute of Medicine Summarizing Population Health: Directions for the Development and Application of Population Metrics. Washington, DC: The National Academies Press; 1998. [Google Scholar]

35. Horsman J, Furlong W, Feeny D, Torrance G. The Health Utilities Index (HUI): concepts, measurement properties and applications. Health Qual Life Outcomes. 2003;1:54. doi: 10.1186/1477-7525-1-54 [PMC free article] [PubMed] [CrossRef] [Google Scholar]

36. Organisation for Economic Co-operation and Development Health at a glance 2017: OECD indicators. http://www.oecd.org/health/health-at-a-glance.htm. Published November 10, 2017. Accessed February 11, 2018.

What is affinity grouping quizlet?

Affinity grouping. determines which things go together. Clustering. segments a heterogeneous population of records into a number of more homogeneous subgroups.

What is classification analysis quizlet?

Classification analysis: is the process of organizing data into categories or groups for its most effective and efficient use. For example, groups of political affiliation and charity donors.

What is another term for data that is already contained in a database or a spreadsheet?

Structured data consists of information already managed by the organization in databases and spreadsheets; it is frequently numeric in nature.