Reduction of the Quality of Contraceptive Counseling (QCC) scale to a short version (QCC-10) in Ethiopia, India, and Mexico

Objective: To reduce the Quality of Contraceptive Counseling (QCC) scale to a shortened version, coined the QCC-10, for use in measuring client-reported quality of counseling across varied settings. Study design: Secondary psychometric analysis of data collected for validating full versions of the QCC scale (QCC-Mexico, QCC-Ethiopia, QCC-India) and expert voting to reduce the original 26 QCC items to a 10-item set. Results: Exploratory factor analysis revealed a clear one-factor solution for the QCC-10 in each country. Factor loadings were consistently > 0.4 for all but two items; both were retained due to their importance to content validity. Internal consistency reliability of the QCC-10 was α = 0.8 in Mexico and Ethiopia, and α = 0.5 in India. QCC-10 scores were highly and positively correlated with a dichotomous overall measure of client experience and intention to initiate selected method, indicating convergent validity. Conclusion: The QCC-10 offers an innovative, cross-cultural approach to measuring quality in contracep- tive counseling. Future efforts should examine its validity and reliability for use globally, with additional exploration of how to best measure negative aspects of care, particularly in India where such items were problematic. Thoughtful, nuanced measurement of client perspectives on their counseling experiences, available via the QCC-10, is critical to monitoring and improving quality of person-centered care and the fulﬁlment of human rights in contraceptive services worldwide. Implications: Cross-cultural, person-centered measures of quality in contraceptive counseling, such as the QCC-10, can help inform efforts to improve quality of family planning services and fulﬁllment of human rights. Future work will continue to explore the validity of this 10-item measure for use in various settings. © 2022 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ )

human rights, including respect for client autonomy ( disrespect and abuse ). The QCC scale was originally developed in Mexico [9] and then adapted and validated for use in Ethiopia and India [10] . Extensive formative research was conducted in these countries to ensure items were grounded in women's preferences for counseling [11][12][13] .
The three validated QCC scale versions (QCC-Mexico, QCC-Ethiopia, QCC-India) have 22 to 26 items, depending on the context [ 9 , 10 ]. These original versions of the scale are useful tools for research that aims to explore nuances in quality of contraceptive counseling. At the same time, a shorter version of the scale is needed for wider use, especially for integration with data collection mechanisms that are unlikely to accommodate the full scale. A shortened version of the QCC scale may prove particularly beneficial for use within routine data systems in health facilities as part of continuous quality improvement processes and for efforts to compare counseling experiences across different geographies.
This paper describes the process used to reduce the QCC scale to a shortened version, coined the QCC-10, for application in contraception programs that seek to assess and monitor clientreported quality of counseling.

Data source
We conducted psychometric analysis of client survey data collected previously for purposes of validating the original QCC scales (QCC-Mexico, QCC-Ethiopia, QCC-India) and triangulated findings across countries to reduce the QCC scale to 10 items. The study in Mexico ( N = 499) was conducted in public clinics and hospitals in two states (San Luis Potosí and Mexico City), including 129 different providers, and was approved by the Harvard T.H. Chan School of Public Health Institutional Review Board (IRB) [9] . The study in Ethiopia ( N = 599) was conducted in Addis Ababa in public health centers and franchise clinics of a large non-governmental organization providing reproductive health services (number of providers included was not documented) and was approved by the University of California, San Francisco (UCSF) and St. Paul's Hospital Millennium Medical College IRBs [10] . The study in India ( N = 313) was conducted in government public health centers in Gujarat state (number of providers included was not documented) and was approved by the Centre for Operations Research and Training (CORT) IRB and deemed exempt by the UCSF IRB due to the lack of identifying information provided to UCSF researchers from the CORT team [10] . All surveys were administered to clients by study staff.
Eligibility criteria included being female and having spoken with a provider about starting, changing, or discontinuing contraception either on the day of recruitment (Ethiopia and Mexico) or within the past 2 weeks (India). Recognizing that client report of counseling quality is likely clustered by provider, we recruited from a convenience sample within multiple health care settings in each country to encourage variability in client experiences. Demographics of study participants are available in the original validation study manuscripts [ 9 , 10 ]. In summary, participants across all settings were on average in their late 20s and dedicating themselves primarily to household work or childcare. In Mexico and Ethiopia only, the majority had finished secondary school.

QCC scale overview
The QCC scale was developed in Spanish in Mexico [9] and later adapted for Ethiopia (Amharic language) and India (Gujarati language) [10] , drawing on qualitative research on women's experiences with and preferences for contraceptive counseling conducted in each setting [11][12][13] . The country-specific number of QCC scale items ranged from 22 to 26 ( Table 1 ). Most items were consistent across settings ( n = 18); four were unique to India and Ethiopia, four unique to Mexico and Ethiopia, and one unique to India. Items mapped to one of three QCC constructs, including (1) information exchange , or assessment of client needs and provision of information essential for decision-making; (2) interpersonal relations , or relationship-building aspects of counseling (e.g., privacy, confidentiality, trust); or (3) disrespect and abuse , defined as coercion, pressure, or mistreatment by providers. This dimensionality held across all three settings.
The scale was administered with a four-point response scale with response categories for positively worded items ranging from completely agree to completely disagree, and for negatively worded items "yes," "yes with doubts," "no with doubts," and "no." Composite scores and subs-scale scores were calculated using a mean of all item responses. The study team chose mean scoring given the clinical utility of the QCC scale and the need to prioritize ease of calculation of scores in quality improvement contexts.

Item reduction process
Our item reduction process consisted of triangulating quantitative psychometric findings with qualitative consideration of item interpretability (as assessed by expert opinions of study team members representing the three countries) and coverage of the full QCC measurement construct [6] . Our first step was to tally votes among the five co-investigators from the three QCC teams in Mexico, Ethiopia, and India (XQ, IZ, BU, KL, EG) who were asked to vote for 10 priority items from the perspective of item interpretability and content validity, drawing on their knowledge of cultural contexts, participant understanding during piloting and data collection, and alignment with the measurement framework. We then added items in order of number of votes received into a spreadsheet, along with information on factor loadings from prior studies and labels for domain of the QCC construct (information exchange, interpersonal relationship, or disrespect and abuse) covered.
We decided a priori to select a total of no more than 10 items for the short scale to ensure scale length could feasibility be integrated with larger surveys and existing data collection tools, and to retain at least two items for each of the three domains of the QCC construct to ensure coverage of the entire construct. Our second step was to generate an initial list of potential items to retain, including those that either had four or five votes from the five coinvestigators or high ( > 0.4) factor loadings from prior studies, and excluding those with zero co-Investigator votes. Given our desire to develop a scale that could measure clients' counseling experiences regardless of whether an individual decides to use a method, we also excluded items that focus exclusively on information about discontinuation and switching as those are less relevant for people who have not initiated method use. In our final step, we sought to remove additional items that were similar in content to other retained items or in order to reach our target number of 10 items.

Descriptive and psychometric analysis of reduced scale
We examined scale and item properties of the QCC-10 by country to examine validity and reliability across settings. We calculated mean item and scale scores to describe score distributions. We conducted exploratory factor analysis (EFA) to identify the factor structure of the reduced scale, identifying factors with an adjusted eigenvalue of at least one through parallel analysis. We assessed factor loadings once factor structure was determined. We calculated item-rest correlations to examine each item's association with the score of the other items, and Cronbach's alpha to examine internal consistency reliability of the scale. Because four    ( Table 2 ). b Co-Investigators were asked to choose top 10; 5 people voted c In the original validation study, EFA was run on Mexico City data and CFA on San Luis Potosi data; the range of loadings from both analyses is displayed here d This item was added to the QCC Scale for Ethiopia and India based on formative research suggesting this is an important element of quality related to Information Exchange; this item was not fielded in Mexico. In India the wording was modified for clarity, based on cognitive interviews, to read "The provider asked me personal questions …". e Removed due to low factor loadings in CFA in India f This item was added to the QCC Scale for Ethiopia and India based on formative research suggesting this is an important element of quality related to Interpersonal Relationships; this item was not fielded in Mexico g In India, this item was replaced with "I felt scolded because of the way I was using contraception" based on formative research suggesting the importance of social norms around what contraceptives are acceptable to be used depending on the number of children someone has and the greater salience of this potential dimension of discrimination compared to age discrimination h This item was added to the QCC Scale for Ethiopia and India based on formative research suggesting marital status as an important dimension of potential discrimination; this item was not fielded in Mexico of the items selected for the final scale were not fielded in Mexico, our ability to examine psychometric properties of the scale in Mexico is limited; nonetheless we report on findings using the six available items from the QCC-10 that were fielded in Mexico. All analyses were conducted by country and using Stata version 15.

Convergent validity assessment
We assessed convergent validity (i.e., the degree to which QCC-10 scores correlate with other, similar measures, as expected) using an overall measure of client experience with the provider (dichotomized to highest rating vs all else) and a measure of whether the client planned to use or continue using the selected contraceptive method, among women who reported selecting or already using contraception on the day of interview. Analyses were conducted using bivariate logistic regression models, accounting for clustering (by provider in Mexico and site in India and Ethiopia where provider identifiers were unavailable) with robust standard errors. Table 1 displays all 26 QCC items, with investigator votes, factor loadings from prior validation studies, and descriptions of the rationale for selecting or excluding each item for the QCC-10 fi-nal item set. Items received between 0 and 5 votes (mean = 1.9, standard deviation = 1.3); only one item received five votes (prov_insist: "The health care provider pressured me to use the method they wanted me to use").

Item reduction
Eighteen items were considered for retention either because they had four or five votes from the five co-investigators ( n = 3) or high ( > 0.4) factor loadings from prior studies in combination with at least one vote from a co-investigator ( n = 15). We excluded three items from the latter category that focus exclusively on information for individuals who select a method during counseling (method_fail, method_react, and method_stop). From the 15 remaining items, we excluded five deemed not essential for coverage of the QCC construct (STI_info because of the focus on STIs; prov_health, prov_opinion, and prov_listens due to their similarity to other items; and prov_sexlife due to its length). The final 10 QCC-10 items are presented in Table 2 . Table 3 presents item and scale properties of the QCC-10. As previously reported, item means ranged from 3.3-3.9 (out of 4) in Mexico, 2.4 to 4.0 in Ethiopia, and 3.0 to 3.9 in India. Overall QCC-10 scores ranged from 3.1 in Ethiopia to 3.5 in Mexico (six items only). Cronbach's alpha was 0.8 in Mexico and Ethiopia, indicating high reliability for a short scale [14] , while the alpha was only 0.5 in India. Missing data ranged from 10 per item in Mexico to 21 in Ethiopia; no data were missing in India.    a Missing data ranged from 10 in Mexico to 21 in Ethiopia in psychometric analyses; no data were missing in India b A sensitivity analysis examining scale properties without the scold_marital item revealed no change in alpha or factor loadings for other items in India or Ethiopia. Removing both scold_marital and prov_insist did not change item or scale properties in Ethiopia, but improved the alpha in India to be comparable to Ethiopia and Mexico (0.8) and increased the info factor loading in India to 0.5.

ARTICLE IN PRESS
EFA revealed a clear one-factor solution in each country, as only one factor in each country had an adjusted eigenvalue greater than one in parallel analysis (Appendix). Factor loadings were consistently above 0.4 for all but two items: prov_insist and scold_marital . The prov_insist item, examining provider pressure to use a specific method, loaded 0.2 in Mexico, 0.1 in Ethiopia, and met the 0.4 threshold in India. Similarly, the scold_marital item, reflecting biased counseling against client's pursuit of contraceptive methods based on their marital status, loaded 0.3 in Ethiopia and 0.1 in India; the item was not fielded in Mexico. Item-rest correlations were also low ( < 0.4) for these two items. Because these two items capture key dimensions of negative counseling experiences, we retained them to preserve content validity of the scale in line with the original measurement framework [6] . However, due to the low alpha in India, we conducted a sensitivity analysis removing both scold_marital and prov_insist; this improved the alpha in India to be comparable to Ethiopia and Mexico (0.8) and increased the info factor loading to 0.5. We also conducted a sensitivity analysis to evaluate if item and scale properties would improve if just scold_marital was removed; results indicated that item loadings and Cronbach's alpha did not substantially change ( Table 3 ). Table 4 displays results from convergent validity analyses. QCC-10 scores were highly correlated with a dichotomous overall measure of client experience in all countries (Mexico: OR = 10.6, 95% CI = 5.7 −19.6; Ethiopia: OR = 45.5, 95% CI = 23.6 −87.7; India: OR = 4.4, 95% CI = 1.8 −11.0). Note that the confidence inter-vals for these analyses are large likely due to clustering of standard errors, particularly in Ethiopia and India where data were collected at the site rather than provider level. We conducted a sensitivity analysis to examine the relationship of an 8-item version of the scale, removing the two negatively worded items that capture disrespect and abuse ( prov_insist and scold_marital ) and that were less commonly endorsed among participants, and found scores remained highly correlated with the client experience measure (Mexico: OR = 7.0, 95% CI = 4.2-11.8; Ethiopia: OR = 30.7, 95%CI = 15.4-60.9; India: OR = 3.2, 95% CI = 1.5-6.7).

Convergent validity
Among participants who selected a method after their contraceptive counseling visit, there was also correlation between QCC-10 scores and clients' reported intention to initiate the method (Mexico: OR = 3.3, 95% CI = 1.7-6.4; Ethiopia: OR = 3.5, 95% CI = 2.5-4.9). A sensitivity analysis removing prov_insist and scold_marital did not meaningfully change the finding (Mexico: OR = 2.4, 95% CI = 1.3-4.2; Ethiopia: OR = 2.3, 95% CI = 1.6-3.2). We did not conduct this analysis for India, as participants initiated a method by the interview given our recruitment approach, conducting interviews within 2 weeks after the consultation.

Discussion
This study describes the process used to reduce 26 items of the multidimensional Quality of Contraceptive Counseling (QCC) scale, measuring client experience of contraceptive counseling, to a unidimensional 10-item scale using data from contraception clients in Mexico, Ethiopia, and India. The QCC-10 offers a new approach to measuring client experience and fulfilment of human rights in Table 4 Convergent validity results from logistic regression predicting the odds of related variables based on continuous Quality of Contraceptive Counseling short scale Missing data ranged from 1 in Ethiopia to 3 in Mexico; no data were missing in India. We conducted a sensitivity analysis to examine the relationship of an 8-item version of the scale, removing the two negatively worded items that capture disrespect and abuse (prov_insist and scold_marital) and that were less commonly endorsed among participants, and found scores remained highly correlated with the client experience measure (Mexico: OR = 7.0, 95% CI = 4.2-11.8; Ethiopia: OR = 30.7, 95%CI = 15.4-60.9; India: OR = 3.2, 95% CI = 1.5-6.7). 3 We did not examine this outcome in India because participants were interviewed within two weeks of their visit, with those initiating method use prioritized for recruitment. Missing data ranged from 104 in Mexico to 122 in Ethiopia due to some women not having selected a method during their visit in Mexico and Ethiopia. A sensitivity analysis removing prov_insist and scold_marital did not meaningfully change the finding (Mexico: OR = 2.4, 95% CI = 1.3-4.2; Ethiopia: OR = 2.3, 95% CI = 1.6-3.2).

ARTICLE IN PRESS
JID: CON [m5G; December 10, 2022;14:12 ] contraceptive counseling that is appropriate for multiple cultural contexts, including geographies in Latin America, South Asia, and sub-Saharan Africa. The scale's strength lies in the robust formative research conducted to develop the original items [11][12][13] and the three-continent survey data leveraged for the scale reduction. The original QCC-Mexico, QCC-Ethiopia, and QCC-India scales are multidimensional, comprising three subscales: information exchange, interpersonal relationship, and disrespect and abuse. The QCC-10 is a shorter, unidimensional scale that retains items from these three subscales (five, three, and two items for each subscale, respectively) and includes content coverage of the original measurement framework [6] . The QCC-10 retained the high internal consistency reliability observed in the QCC-Mexico and QCC-Ethiopia validation studies, even with the lack of data for four items in Mexico. However, Cronbach's alpha was notably lower among women in India (alpha = 0.5, not typically considered acceptable) [14] . Despite clear evidence from EFA that the QCC-10 in India was unidimensional, this low alpha likely indicates heterogeneity in what individual scale items measure. The two items reflecting disrespect and abuse ( prov_insist and scold_marital ) had the lowest factor loadings in India. We hypothesize these items are responsible for the lower alpha, possibly because they capture a different aspect of quality-focused on understudied, negative experiences-relative to the positively-worded items in the information exchange and interpersonal relationship domains, which loaded 0.4 or higher. Indeed, the removal of the two items in India increased the scale reliability (alpha = 0.8; Table 3 ). The poor performance of the QCC-10 in India suggests the need for further exploration of its suitability in this context. We recommend that future work examine the performance of the prov_insist and scold_marital items in a larger, more diverse sample from India. Given the importance of the provider bias and discrimination items from a content validity perspective, including positive and negative dimensions of care in measures such as the QCC-10 is central to comprehensive assessment of high-quality contraceptive services.
Convergent validity findings offer promising evidence for the QCC-10's construct validity with higher scale scores correlating with anticipated client outcomes (e.g., overall measure of client experience and intention to use/continue method use) across settings. Future research should examine the transferability and psychometric properties of the QCC-10 in other low-and middleincome countries to examine the validity and reliability of the QCC-10 in other contexts and assess if our observed measurement challenges are unique to the Indian study context or exist across geographies. We also recommend future validation studies include nonresponse analyses to examine the minimum number of items required to retain psychometric properties. Further, researchers may choose to explore other ways to score the QCC-10 in their work. Mean scoring was chosen by the QCC project team for the original long-form validation studies and for this reduction study, given the prioritization of facilitating QCC-10 score calculations in clinical contexts. In contexts where more complex score calculation is possible, however, other approaches may be valuable to explore.
The QCC-10 complements other recently developed measures of quality in contraceptive counseling, including the Interpersonal Quality in Family Planning (IQFP) scale (developed and validated in the United States and validated for use in India [ 15 , 16 ]) and its reduced version, the Person-Centered Contraceptive Counseling (PCCC) scale [17] ; and the Method Information Index (MII) and MII-Plus, developed from existing survey items used for decades as part of the Demographic Health Surveys [18] . The IQFP is an 11-item scale similar in length to the QCC-10, focuses on personcenteredness in contraceptive counseling, yet, the QCC-10 is distinct in the scope of the construct measured; specifically the QCC-10 comprises domains of privacy, confidentiality, and nondiscrimination not covered in the IQFP and includes items probing for negative experiences, such as pressure to use contraception. While the PCCC and MII are both shorter measures that can be easily integrated into surveys, compared to the IQFP and QCC-10, they also offer less nuance in measuring quality of counseling. Similarly, with simple yes/no response options and a focus on counseling content, not experience, MII/MII-Plus data result in less response variation than other available measures, which utilize Likert scales for responses; such homogeneity has proven challenging from a research perspective. An important and foundational benefit of the QCC-10 relative to these other measures is the robust, formative research and cognitive testing that was conducted in the three settings during the item refinement phase; such effort s were critical to ensuring content validity in each new setting before quantitatively testing the items [9][10][11][12][13] .
This study was limited by the narrow geographic scopes included in each of the original QCC validation studies (two states in Mexico, the capital city of Ethiopia, and one state in India). Despite this, the QCC-10 is unique in that it was developed with data from three diverse settings. Additionally, as part of the reduction process, we prioritized information exchange items relevant to selection of a method and excluded several items that focus exclusively on information about discontinuation and switching among contraceptive method users, given our desire to develop a scale that could measure clients' counseling experiences regardless of whether they decide to use a method. Thus, the resulting QCC-10, compared to the longer, country-specific QCC scales, covers a limited scope of the entire information exchange component of the original measurement framework [6] . Finally, the data sources used for Ethiopia and India do not include provider identifiers and, thus, we are unable to account for clustering of women's experiences by provider in our Ethiopia and India analyses.
The QCC-10 offers an innovative, cross-cultural approach to measuring quality in contraceptive counseling. Future efforts should examine its validity and reliability for use globally, with additional exploration of how to best measure negative aspects of care in India where the QCC-10 Disrespect and Abuse items were problematic. Thoughtful, nuanced measurement of client perspectives on the counseling they receive, available via the QCC-10, is critical to monitoring and improving quality of care and the fulfilment of human rights in contraceptive services worldwide.

Data availability
Data will be available on Dryad.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Funding disclosures
This work was supported by the David and Lucile Packard Foundation [Grant number: #2018-67680 ].