Conjoint analyses of patients’ preferences for primary care: a systematic review

Lim, Audrey Huili; Ng, Sock Wen; Teh, Xin Rou; Ong, Su Miin; Sivasampu, Sheamini; Lim, Ka Keat

doi:10.1186/s12875-022-01822-8

BMC Primary Care

Table 1 Characteristics of included studies (N = 35)

From: Conjoint analyses of patients’ preferences for primary care: a systematic review

Characteristics			Characteristics
*Study settings*	n	%	*Questionnaire design*	n	%
Publication year			Choice contexts^a
2010–2022	21	60.0	Choosing primary care for self (not specified)	31	88.6
1997–2009	14	40.0	Choosing primary care for self when the current one closes	3	8.6
			Choosing primary care for self after moving to a new city	1	2.9
Continent			Choosing primary care for a friend / relative	1	2.9
Europe	23	65.7
Asia	5	14.3	Types of visits^a
North America	4	11.4	Acute: minor	19	54.3
Australia & New Zealand	2	5.7	Non-specific / other^c	16	45.7
Africa	1	2.9	Chronic	6	17.1
			Acute: major	4	11.4
Country’s income level^b
High income	29	82.9	Types of attributes^a
Low & middle income	6	17.1	Process	33	94.3
			Outcomes	32	91.4
Sources of funding			Structure	18	51.4
Government	16	45.7
Not reported	10	28.6	Methods to identify attributes & levels^a
Independent organization	5	14.3	Literature review	25	71.4
Academic institution	4	11.4	Qualitative research	22	62.9
			Not reported	4	11.4
Study samples^d	Mean	SE	Policy	3	8.6
Sample size	881.8	739.3	Others	3	8.6
Response rate (%)	62.8	22.9	Expert opinion	2	5.7
Age	51.6	8.7
Percentage of men (%)	41.9	8.7	Factors affecting preference heterogeneitya^e
			Did not examine any factor	19	54.3
Type of conjoint analysis	n	%	Predisposing characteristics	10	28.6
Choice-based	33	94.3	Enabling resources	9	25.7
Rating-based	2	5.7	Needs	5	14.3
			Health behaviour	2	5.7
*Study design*	n	%
Recruitment setting			Methods to generate choice set
Primary care facilities	19	54.3	Software	17	48.6
Community	15	42.9	Not reported	16	45.7
Not reported	1	2.9	Catalogue	2	5.7
Survey administration			Reported design efficiency^f
Self-completed	22	62.9	D-efficient	19	54.3
Interviewer administered	7	20.0	Not reported	16	45.7
Computerized interview	3	8.6
Computer aided telephone	2	5.7	*Study quality*^g	n	%
interview			Main analysis
Self-completed & Interviewer administered	1	2.9	High	29	82.8
			Low	6	17.1
*Study analyses*	n	%	Sensitivity analysis
Statistical models^a			High	25	71.4
Logit	26	74.5	Low	10	28.6
Probit	8	22.9
Latent class analyses	3	8.6
Others^h	2	5.7

^a Sums to > 100% as a study may report / examine more than one of these characteristics
^b Categorised based on The World Bank classification on 21 May 2021 at (https://datahelpdesk.worldbank.org/knowledgebase/articles/906519-world-bank-country-and-lending-groups)
^c Six studies specified other reasons for visits e.g., general consultation, annual check-up, and appointments for other family members. The remaining nine studies did not specify the reason for visits
^d Not all studies reported all study characteristics: all 35/35 studies reported sample size, 24/35 reported response rate, 21/35 reported mean age of respondents, 31/35 studies reported percentage of men
^e The factors were based on the Anderson model of healthcare utilization, which categorizes factors affecting healthcare utilization into predisposing characteristics (e.g., age), enabling resources (e.g., income level), needs (e.g., health status) and health behavior (e.g., utilization of healthcare)
^f D-efficiency score indicates the extent to which the studies are balanced and orthogonal. Perfectly efficient designs are balanced (each level appears equally often within an attribute) and orthogonal (each pair of levels appears equally often across all pairs of attributes within the design)
^g Based on the Conjoint Analysis Applications in Health – a Checklist: A Report of the ISPOR Good Research Practices for Conjoint Analysis Task Force. In the main analysis, studies that fulfilled all the items in the checklist (either partially or completely) were considered of acceptable quality. In sensitivity analysis, only studies that completely fulfilled at least 5 items out of 10 in the checklist were considered of acceptable quality
^h Other statistical models are Hierarchical Bayes estimation and fractional replication methodology in the Categories module of SPSS

Back to article page

ISSN: 2731-4553

Contact us

General enquiries: journalsubmissions@springernature.com