Table 3

Methods

Author/Methods	Van den Akker et al, 2001, 2006 ‘Registration Network Family Practices (RNH) Cohort’	Schellevis et al, 1994	Bayliss et al, 2004	Perkins et al, 2004	Parkerson et al, 1995
Location/s	South Netherlands	South-east Netherlands	US: Boston, MA; Chicago, IL; and Los Angeles, CA	US: Indiana state	US: Caswell County, NC

Primary care settings recruited	15 practices	7 practices	3 prepaid group practices (one from one HMO in each city) plus 23 of their satellite facilities; 25 multispecialty group facilities	1 practice	1 practice

Selection of primary care setting/s	Non random — 42 GPs in 15 practices registered with the RNH database.	Non random — Seven practices (15 GPs) from 103 that had taken part in a previous study selected based on them having participated in the previous study during a specific time period and the practice being located in the south-east of the Netherlands.	Non random — Selected study sites (cities) based on the size of the HMO, having both prepaid and fees-for-service arrangements, number of physicians, and willingness to take part. Three study sites met the criteria. In each city, five or six practice sites were sampled from each group practice HMO. Populations of clinicians were sampled according to specialty training, age and experience.	Not stated	Not stated

Number of GPs approached	n/a	18	1791	Not stated	Not stated

Number of GPs participating	42	15	225 contributed patients	Not stated	2 GPs & 2 general internists

Patient population studied	Sample at follow-up — mean age 52.4 years (SD 16.8), 22.3% of follow-up sample did not have any disease at the start of study, 21.4% had one disease, and 56.3% had two or more diseases	Hypertension, n = 549 (ages 22–92, 86.3% with single disease); chronic ischaemic heart disease, n = 183 (ages 34–97, 73.8% with single disease); diabetes mellitus, 119 (ages 29–88, 68.9% with single disease); CRD, n = 252 (ages 3–86, 88.9% with single disease); and osteoarthritis knee and/or hip, n = 80 (ages 39–87, 68.8% with single disease). Sample included people with one chronic disease and those with two or more.	Mean age 57.6 (SD 15.4); 58.3% married; 46.4% employed; 19.3% at or below 200% of poverty level; 82.5% white (versus non-white). 14.6% education less than high school; 28.5% high school graduate; 28.5% greater than high school; 12.1% college graduate; 16.3% greater than college; mean number of main diseasesa 1.5; mean number of additional diseasesb 0.7; hypertension alone (referent group) 0.3; remaining subjects 0.8.	Vulnerable older adults, average age was 68.9 years; 56% were African American. Approximately 75% of the patients had Medicare and 27% had Medicaid; 20% patients were smokers and 45% patients were obese. Patients averaged 7 medications, 6.4 unique pharmacy subclasses, and 5.2 chronic disease classes; and averaged eight ambulatory visits per year.	Ambulatory primary care patients aged 18–65 years

Number of patients approached	6113	Practice population of 23 534 whose records were searched or who were identified through a visit to /contact with the practice in first 3 months of study (identified by GPs).	3589 (eligible for longitudinal study out of 28 257 from which both a longitudinal and cross-sectional study would be recruited).	3601	561

Number of patients agreeing to participate	Not stated	Not stated	3589	3496	534 (95.2%)

Number of patients in study at baseline/ with baseline data collected	3745	962	2235	3496	414 (74%)

Patients sampled from where?	RNH database	GP	From MOS cohort study of primary care patients selected via HMOs ‘facilities’	Primary care practice	Primary care practice

How were patients sampled? (consecutive patients, mail out from practice list, self-reported, from electronic records)	Random sample of patients stratified to number of disease prior to study.	GPs identified from records (or through a visit to /contact with the practice in first 3months of study) all patients for the study group. Control group selected using data from the Dutch National Survey of general practice.	Convenience sample — patients approached during a visit with a MOS clinician during a 2-week period in Feb 1986. Adults who visited healthcare provider in 9-day period starting in Feb 1986.	Convenience sample — patients aged ≥60 with a scheduled primary care appointment between 15 July 1999 and 31 August 2001, were eligible ending April 1991.	Convenience sample (12 age-sex-race categories) — patients approached by research assistant when presenting at clinic during 8-month period.

Selection of patients	Random	Non random	Non random	Non random	Non random

Method of randomisation	Not stated	Not applicable	Not applicable	Not applicable	Not applicable

Exclusion criteria (for patients)	Illiterate, spoke little or no Dutch, incapable of participating due to mental or physical status.	Patients who left the practice during the 21-month study period & patients who had received follow-up care from a specialist before the start of the study.	Cannot read English	Not stated	Illiterate, too sick to participate

Criteria for inclusion in cohort	Those on the RNH database aged ≥20. Cases were defined as subjects who had new multimorbidity, that is two or more new disorders registered on the problem list within a period of 3 years (1 September 1992–31 August 1995). Controls — no new disease in selection period. Additional control group — one new disease registered during selection period.	Diagnosis of at least one of the following diseases: hypertension, chronic ischaemic heart disease, diabetes mellitus, CRD (asthma, chronic bronchitis, emphysema), and osteoarthritis of knee and/or hip. Diagnosis made before 1 January 1988, diagnosis in agreement with the diagnostic inclusion criteria of ICHPPC-2-and patient not receiving follow-up care for the disease from a specialist at the start of the study.	Diagnosis with one or more of diabetes, CAD, CHF, CRD, musculoskeletal conditions, and depression. Referent population with a diagnosis of hypertension exclusive of any of the other major comorbid conditions. Completion of baseline data collection.	Community-dwelling patients ≥60 years of age with a scheduled primary care appointment between 15 July 1999 and 31 August 2001, were eligible.	Presenting to clinic for health care visit during 8-month period ending April 1991; age 18–65 years

Patient screening procedures	Searches of an electronic database.	Patient attendance at the practice during a specific time period.	Physician reports verified by study clinical staff and through a patient questionnaire.	Patient attendance at the practice during a specific time period.	Patient attendance at the practice during a specific time period.

Final cohort size (% female)	3551 (49.3%)a (2001), 3460 (49.8%)b (2006)	962 (overall not stated but 53.8%)	1574 (58.7%)	3496 (69%)	413 (58.6%)

Rationale for sample size	Not stated	Not stated	Not stated	Not stated	Not stated

Percentage of patients lost to follow-up overall	7.6% (285 out of 3745)	0%	29.6%	0%	0.2% (1 person)

Percentage of non-responders	5.2% could not be matched on database at follow-up (194 out of 3745).	Not applicable	Not described separately from drop outs.	0%	9% (50/534) illiterate and could not complete questionnaire at baseline.

Percentage dropped out of cohort	2.4% died (91/3745)	n/a	29.6% (661/2235)	0%	0.2% (1 person)

Retention rate	92.4%	100%	70.4%	100%	99.8%

Missing data	c. 40% questionnaires with missing data at baseline data collection — exact % not given.	c. 30% of consultations not recorded on research forms.	Not applicable	0%	7.9% (42/534) of patients consenting had excessive data missing from intake questionnaire.

Characteristics of the study subjects lost to follow-up	Not stated	Not applicable	Were younger and had lower income: no differences in initial health status.	Not applicable	Not stated

Reasons for drop out/loss to follow-up	91 died, for 194 data could not be matched on database at follow-up.	Not applicable	Variety of reasons including refusals and failure to contact (n = 661; 29.6%); 137 (6.1%) who died during follow-up were included in the analysis.	Not applicable	Records could not be found for 1 person at audit so incomplete follow-up data available.

Number of patients in cohort with multimorbidity (at follow-up unless stated otherwise)	n = 305 (8.6%)a (2001); n = 290 (8.4%)b (2006) had new multimorbidity.	n = 268 (22.7%) had ≥2 chronic diseases at baseline.	n = 686 (43.6%)	Not stated	216

Total length of follow-up period	2 years	21 months	4 years	1 year	18 months

Number of follow-up points	Continuous	Continuous	1	Continuous	Continuous

Timing of each follow-up	2 years (continuous over period)	21 months (continuous over period)	4 years	1 year (continuous over period)	18 months (continuous over period)

Study period	January 1996/December 1997 (baseline) to January 1998/December 1999 (follow-up).	January 1988–October 1989	1986–1990	1999–2002 (followed for 1 year after patient index visit).	September 1990/April 1991 till February/September 1992 (followed for 18 months after recruitment).

Baseline data collection methods	Electronic record extraction/ patient report	Doctor report	Patient report/doctor report	Electronic record extraction	Patient report/doctor report/ record extraction

Baseline measurements	BMI (body mass index), smoking, alcohol use, sports, family medical history, long-term difficulties, life events, health locus of control, coping style, social network, values, morbidity and multimorbidity of chronic, recurrent and high impact diseases	Chronic diseases diagnosed from list of five diseases	Diagnoses; physical health status	Diagnostic ICD-9 codes, ambulatory visits and inpatient stays for year prior to index visit and all prescription medications for prior year. ACG categories & CCI. Calculated CDS, one to predict healthcare costs (CDS-HC) and one to predict mortality (CDS-M), to identify chronic disease class, then rated by physicians for likely healthcare utilisation and mortality over 1 year.	Health status and severity of illness

Self-reported or clinically determined diagnosis?	Clinically determined	Clinically determined	Clinically determined and self-reported	Clinically determined	Clinically determined

Outcomes data collection sources	Electronic record extraction/ patient report	Doctor report	Patient report/doctor report	Electronic record extraction	Doctor report/record extraction

Outcomes measurements	New multimorbidity, new morbidity 2001); International Classification of Primary Care (ICPC) codes representing a diagnosis for new diagnoses (2006).	Number/rate of consultations and episodes of disease, incidence rate and nature of ‘intercurrent’ morbidity (other new illnesses including acute ones).	Analysed categorical change (worse versus same/better) in Medical Outcomes Study SF-36 physical component summary (PCS) scores. Categorical change defined as a change of ≥6.5 points in PCS. Also assessed linear change in PCS scores.	All outpatient and inpatient visits and medical charges and death certificate information.	Frequency of primary care visits, severity of illness, health care charges, inpatient stays/ care charges, inpatient stays/ referrals, follow-up severity of illness, utilisation of office and referral services, cost of office health care during 18-month period.

Data collection methods/ tools	Patient reported: baseline data collected by patient self-completion postal questionnaire including: long-term difficulties questionnaire; VRMG — Recent events ; questionnaire Multidimensional health locus of control; Utrecht coping list. Record extraction: RNH database — initial data on morbidity/ multimorbidity and follow-up data on new morbidity and new multimorbidity	Doctor reported: recorded all consultations with cohort on special research forms; recorded ≥1 diagnosis at the highest diagnostic level appropriate, whether the episode was new or pre-existing.	Patient reported: SF36 Health Survey. Doctor reported: diagnoses confirmed by independent clinical exam by MOS staff.	Record extraction: using Regenstrief Medical Record system (RMRS) electronic medical record system for baseline and outcome measures. Diagnostic ICD-9 codes, ambulatory visits, inpatient stays, and all prescription medications from pharmacy database. Calculated ACG categories using outpatient data only and CCI using inpatient data only. Calculated CDS using prescription records.	Patient reported: DUKE (Duke Health Profile) questionnaire a 17-item generic functional health health measure including physical, mental & social health; Doctor reported: of illness using DUSOI; Record extraction: outcomes data — follow-up severity of illness, utilisation of office and referral services, cost of office health care during 18-month period by medical record audit. Audit check of severity of illness by research team to confirm severity rating by GPs.

Definition of ‘multimorbidity’ (or ‘comorbidity’a where this term was used instead)	Multiple diseases (2001), co-occurrence of diseases (2006).	Not stated	Inter-relationships between different diseases and between diseases and age or other health-related sociodemographic variables.	The total burden of illnesses across multiple potential conditions unrelated to the patient's principal or target diagnosis.	Not stated

How was ‘multimorbidity’ (or ‘comorbidity’) operationalised?	Two or more new diseases that are permanent, chronic, recurrent or have lasting consequences.	Included: hypertension, chronic ischaemic heart disease, diabetes mellitus, CRD and osteoarthritis.	Having one of the six chronic conditions of interest: diabetes, CAD, CHF, CRD, musculoskeletal conditions, and depression, plus at least one of 16 (unspecified) additional conditions on top of ‘main diseases’. Chosen based on their high prevalence as well as frequent assessment in the literature oncomorbidity and chronic disease management.	Five measures of multimorbidity used. Number of chronic diseases out of 10 (arthritis, coronary artery disease, cancer, congestive heart failure, COPD, diabetes, hypertension, liver disease, renal disease, and stroke). ACG categories calculated by classifying patients into one of 34 diagnostic groups by clustering similar conditions based services resource. The CCI which assigns weights to 19 conditions, based on their risk of mortality, the weights are then summed for each patient. A CDS based on the number of different chronic diseases as inferred from the subject's prescribed medications.	Two or more health problems on the ICHPPC-2.

Criteria for inclusion in cohort	South Netherlands	Diagnosis of at least one of the following five diseases: hypertension, chronic ischaemic heart disease, diabetes mellitus, CRD, and osteoarthritis of knee and/or hip.Diagnosis made before 1 January 1988, diagnosis in agreement with the diagnostic inclusion criteria of ICHPPC-2-defined and patient not receiving follow-up care for the disease from a specialist at the start of the study.	Diagnosis with one or more of six conditions: diabetes, CAD, CHF, CRD, musculoskeletal conditions, and depression. Referent population with a diagnosis of hypertension exclusive of any of the other major comorbid conditions (but they could have other long-term conditions). Completion of baseline data collection.	Community-dwelling patients ≥60 years with a scheduled primary care appointment between 15 July 1999 and 31 August 2001, were eligible.	Presenting to clinic for health care visit during 8 month period ending April 1991; aged 18–65 years.Two or more health problems on ICHPPC-2

ACG = Ambulatory Care Group. CAD = coronary artery disease. CCI = Charlson Comorbidity Index. CDS = Chronic Disease Score. CHF = coronary heart failure. COPD = chronic obstructive pulmonary disease. CRD = chronic respiratory disease. DUSOI = Duke severity of illness checklist. HMO = health management organisation. ICHPPC-2 = (International Classification of Health Problems in Primary Care-2). PCS = physical component summary. RNH database = Research Networtk Family Practices database. SF-36 = 36-Item Short-Form Health Survey.*includes those who died in sample. bexcludes those who died from sample.