Abstract

Introduction

Every year there are approximately 22 million unsafe abortions in the world, 98% of which in low and middle-income countries ¹1. World Health Organization. Safe abortion: technical and policy guidance for health systems. Geneva: World Health Organization; 2012.. According to the World Health Organization (WHO) ¹1. World Health Organization. Safe abortion: technical and policy guidance for health systems. Geneva: World Health Organization; 2012., one-fourth of these abortions require timely medical care to avoid complications.

In Brazil, abortion is only allowed when the pregnancy results from rape or involves risk to the woman’s life or fetal anencephaly. Yet the legal prohibition does not prevent abortion in practice ²2. Monteiro MFG, Adesse L, Drezett J. Atualização das estimativas da magnitude do aborto induzido, taxas por mil mulheres e razões por 100 nascimentos vivos do aborto induzido por faixa etária e grandes regiões. Brasil, 1995 a 2013. Reprod Clim 2015; 30:11-8.. According to a national survey in urban areas in 2016, 18% of Brazilian women 35 to 39 years of age reported having undergone an abortion ³3. Diniz D, Medeiros M, Madeiro A. Pesquisa Nacional de Aborto 2016. Ciênc Saúde Colet 2017; 22:656-60.. Illegality contributes to unsafe abortions, and the resulting complications lead to more than 200.000 hospitalizations per year ²2. Monteiro MFG, Adesse L, Drezett J. Atualização das estimativas da magnitude do aborto induzido, taxas por mil mulheres e razões por 100 nascimentos vivos do aborto induzido por faixa etária e grandes regiões. Brasil, 1995 a 2013. Reprod Clim 2015; 30:11-8.. Women encounter problems in health services, ranging from difficulty in access to hospital beds to situations of discrimination during hospitalization ⁴4. Araújo TVB, Aquino EML, Menezes GMS, Alves MTSSB, Almeida M-d-CC, Alves SV, et al. Delays in access to care for abortion-related complications: the experience of women in Northeast Brazil. Cad Saúde Pública 2018; 34:e00168116.. Delays in care determine the severity of complications ⁵5. Santana DS, Cecatti JG, Parpinelli MA, Haddad SM, Costa ML, Sousa MH, et al. Severe maternal morbidity due to abortion prospectively identified in a surveillance network in Brazil. Int J Gynaecol Obstet 2012; 119:44-8.. Still, there are few Brazilian studies on quality of post-abortion care ⁶6. Menezes G, Aquino EML. Pesquisa sobre o aborto no Brasil: avanços e desafios para o campo da saúde coletiva. Cad Saúde Pública 2009; 25 Suppl 2:S193-204.^,77. Departamento de Ciência e Tecnologia, Secretaria de Ciência, Tecnologia e Insumos Estratégicos, Ministério da Saúde. Aborto e saúde pública no Brasil: 20 anos. Brasília: Ministério da Saúde, 2009. (Série B. Textos Básicos de Saúde)..

The international literature features health facilities for care of obstetric emergencies, which includes abortion complications ⁸8. Holmer H, Oyerinde K, Meara JG, Gillies R, Liljestrand J, Hagander L. The global met need for emergency obstetric care: a systematic review. BJOG 2015; 122:183-9.. Studies on women’s perceptions of such care are less common and have been conducted in jurisdictions where abortion is legal ⁹9. McLemore MR, Desai S, Freedman L, James EA, Taylor D. Women know best - findings from a thematic analysis of 5,214 surveys of abortion care experience. Womens Health Issues 2014; 24:594-99.. This gap motivated the GravSus-NE study on hospital abortion care in the Brazilian Unified National Health System (SUS) in three state capitals in Northeast Brazil: Salvador (Bahia State), Recife (Pernambuco State), and São Luís (Maranhão State) ¹⁰10. Aquino EM, Menezes G, Barreto-de-Araújo TV, Alves MT, Alves SV, Almeida MdCCd, et al. Qualidade da atenção ao aborto no Sistema Único de Saúde do Nordeste brasileiro: o que dizem as mulheres? Ciênc Saúde Colet 2012; 17:1765-76.. Quality of care was defined on the basis of the ethical and normative framework for women’s comprehensive healthcare and abortion care in particular ¹¹11. Departamento de Ações Programáticas Estratégicas, Secretaria de Atenção à Saúde, Ministério da Saúde. Política nacional de atenção integral à saúde da mulher: plano de ação 2004-2007. Brasília: Ministério da Saúde; 2004. (Série C. Projetos, Programas e Relatórios).^,1212. Secretaria de Atenção à Saúde, Ministério da Saúde. Atenção humanizada ao abortamento: norma técnica. Brasília: Ministério da Saúde; 2005. (Série A. Normas e Manuais Técnicos) (Série Direitos Sexuais e Direitos Reprodutivos - Caderno 4).. Four essential dimensions of care were considered: reception and orientation, technical quality of care, inputs/physical environment, and continuity of care ¹⁰10. Aquino EM, Menezes G, Barreto-de-Araújo TV, Alves MT, Alves SV, Almeida MdCCd, et al. Qualidade da atenção ao aborto no Sistema Único de Saúde do Nordeste brasileiro: o que dizem as mulheres? Ciênc Saúde Colet 2012; 17:1765-76..

Although a publication from 2013 ¹³13. Rocha BNGA, Uchoa SAC. Avaliação da atenção humanizada ao abortamento: um estudo de avaliabilidade. Physis (Rio J.) 2013; 23:109-27. concluded in favor of the evaluability of the model of abortion care proposed by the Brazilian Ministry of Health ¹²12. Secretaria de Atenção à Saúde, Ministério da Saúde. Atenção humanizada ao abortamento: norma técnica. Brasília: Ministério da Saúde; 2005. (Série A. Normas e Manuais Técnicos) (Série Direitos Sexuais e Direitos Reprodutivos - Caderno 4)., at the beginning of this study no instruments were identified to assess quality of care for unsafe abortion from the patients’ perspective, except for a set of items in a WHO document ¹⁴14. World Health Organization. Maternal health and safe motherhood programme. Studying unsafe abortion: a practical guide. Geneva: World Health Organization, 1996.. Since these items partly covered the intended dimensions, they were used as the point of departure for developing an appropriate questionnaire for Brazil’s characteristics and standards for abortion care ¹²12. Secretaria de Atenção à Saúde, Ministério da Saúde. Atenção humanizada ao abortamento: norma técnica. Brasília: Ministério da Saúde; 2005. (Série A. Normas e Manuais Técnicos) (Série Direitos Sexuais e Direitos Reprodutivos - Caderno 4). (hereinafter “QualiAborto-Pt”).

A previous article ¹⁵15. Aquino EML, Menezes GMS, Barreto-de-Araújo TV, Alves MT, Almeida MCC, Alves SV, et al. Avaliação da qualidade da atenção ao aborto: protótipo de questionário para usuárias de serviços de saúde. Cad Saúde Pública 2014; 30:2005-16. discussed the first stage in the development of this questionnaire, involving the formal process of translation and semantic refinement of the original set of WHO items. Questions from other studies were added, and still other questions were developed by our own team. This article continues the process, aimed at assessing the previous prototype’s psychometric properties. In order to fine-tune the questionnaire and propose a more effective and efficient factor solution, the study visits the configural and metric structures of QualiAborto-Pt. We assess the dimensions originally proposed according to the theoretical frame of reference ¹⁰10. Aquino EM, Menezes G, Barreto-de-Araújo TV, Alves MT, Alves SV, Almeida MdCCd, et al. Qualidade da atenção ao aborto no Sistema Único de Saúde do Nordeste brasileiro: o que dizem as mulheres? Ciênc Saúde Colet 2012; 17:1765-76., reliability (discriminant validity), factor specificity, and absence of redundancy in the component items, as well as the set’s discriminant factor validity (among the subscales).

Methods

Study design, sampling procedures, and data production

This cross-sectional study included women 18 years or older living in the municipalities included in the study, hospitalized for abortion or its complications, independently of the clinical severity and reported type (induced or spontaneous abortion). The sample excluded abortions permitted by Brazil’s legislation, resulting from ectopic pregnancy and hydatidiform mole, and those resulting from other abnormal products of conception whose clinical and legal justifications support uterine evacuation under safe conditions.

The sample size was estimated at 2,562 women to compare prevalence of serious complications between the three cities. Based on the literature ¹⁶16. Say L, Pattinson RC, Gülmezoglu AM. WHO systematic review of maternal morbidity and mortality: the prevalence of severe acute maternal morbidity (near miss). Reprod Health 2004; 1:3. and on the cities’ secondary data, we assumed a 100% difference in the prevalence of serious complications in Salvador (p2 = 0.08) compared to Recife (p1 = 0.04). We interviewed the women hospitalized from August 31 to December 30, 2010, for abortion complications in all the public hospitals providing this type of care (seven in Salvador, eight in Recife, and four in São Luís) until we reached the calculated sample size.

Among the 3,071 eligible women, there were 5.9% of losses (due to hospital discharge or death before the interview) and 2.7% of refusals. The 2,804 interviewed women had a median age of 27 years, and 57% had a complete secondary education.

Face-to-face interviews were performed by female health professionals, protected by professional confidentiality, trained for 40 hours, and certified for this purpose. The interviews took place while patients were waiting for hospital discharge, although it was possible to perform the interviews partially before then. Questions on inputs/physical environment and continuity of care could only be answered after the patient had been informed of her discharge. Only 5.3% of the interviewees did not answer this section of post-discharge questions. The population actually analyzed include 2,336 patients.

Data analysis

The first stage included a series of meetings for the selection and refinement of the 55 items on the prototype questionnaire ¹⁵15. Aquino EML, Menezes GMS, Barreto-de-Araújo TV, Alves MT, Almeida MCC, Alves SV, et al. Avaliação da qualidade da atenção ao aborto: protótipo de questionário para usuárias de serviços de saúde. Cad Saúde Pública 2014; 30:2005-16.. The central focus was to assess whether the items should be excluded or maintained, and whether they could be improved. Decisions were also based on preliminary exploratory factor analyses. The process led to a second prototype containing a reduced number of items (details discussed in the Results section).

In the second stage, dimensional scrutiny of this reduced prototype began with the investigation of the original four-dimensional structure: reception and orientation; technical quality of care; inputs/physical environment; and continuity of care ¹⁵15. Aquino EML, Menezes GMS, Barreto-de-Araújo TV, Alves MT, Almeida MCC, Alves SV, et al. Avaliação da qualidade da atenção ao aborto: protótipo de questionário para usuárias de serviços de saúde. Cad Saúde Pública 2014; 30:2005-16.. Confirmatory Factor Analysis (CFA) was used for this purpose ¹⁷17. Brown TA. Confirmatory factor analysis for applied research. New York: The Guilford Press; 2015.^,1818. Loehlin JC. Latent variable models: an introduction to factor, path, and structural equation analysis. Mahwah: Lawrence Erlbaum; 2003..

Anticipating problems in the original dimensionality, the prototype’s configural structure was then submitted to exploratory reassessment. An interim principal components analysis was performed to obtain eigenvalues aimed at orienting the subsequent analyses ¹⁹19. Gorsuch RL. Factor analysis. Hillsdale: Lawrence Erlbaum; 1983.. These consisted of adjusting Exploratory Structural Equation Models (ESEM) with 2 to 6 factors ²⁰20. Marsh HW, Muthén B, Asparouhov A, Lüdtke O, Robitzsch A, Morin AJS, et al. Exploratory structural equation modeling, integrating CFA and EFA: application to students' evaluations of university teaching. Struct Equ Modeling 2009; 16:439-76.. The presence of residual item correlations (error) was also inspected, since violation of local (conditional) dependence may be indicative of item redundancies ¹⁷17. Brown TA. Confirmatory factor analysis for applied research. New York: The Guilford Press; 2015.. All the ESEM used the geomin oblique rotation method ¹⁸18. Loehlin JC. Latent variable models: an introduction to factor, path, and structural equation analysis. Mahwah: Lawrence Erlbaum; 2003.^,2121. Muthén LK, Muthén BO. Mplus user's guide. 5th Ed. Los Angeles: Muthén & Muthén; 2007.. Evaluation of the configural structure followed the theoretical meaning.

Having identified the “best” ESEM, a Corresponding Confirmatory Factor Analysis (CFA) model was adjusted, freely estimating the item loadings pertaining to a given factor, but limiting the other factors to zero. In addition to reassessing the factor loading sizes and possible error correlations, this stage involved the evaluation of Fator-Based Discriminant Validity (FDV). This consisted of comparing the square root of the each factor’s Average Variance Extracted (AVE) with the respective factor correlations (ρvef•vs.ϕf↔κ) ¹⁷17. Brown TA. Confirmatory factor analysis for applied research. New York: The Guilford Press; 2015.^,2222. Kline RB. Principles and practice of structural equation modeling. New York: The Guilford Press; 2015.. AVE relates the quantity of information in a characteristic captured by the (manifest) items to the amount of error in the measurement, that is, the portion not explained by the latent factor ²³23. Hair JF, Black B, Babin B, Anderson RE, Tatham RL. Multivariate data analysis. Englewood Cliffs: Prentice-Hall; 2007. Mathematically, it is a function of the standardized item loadings and the respective residuals ²⁴24. Raykov T. Analytic estimation of standard error and confidence interval for scale reliability. Multivariate Behav Res 2002; 37:89-103.. Values vary from 0 to 1. A violation of FDV was defined as ρvef• < ϕf↔κ in at least one of the factors, as long as statistically significant at 5%. This analysis used the bootstrap method (B = 1,000 replications) ²⁴24. Raykov T. Analytic estimation of standard error and confidence interval for scale reliability. Multivariate Behav Res 2002; 37:89-103.^,2525. Efron B, Tibshirani R. An introduction to the bootstrap. London: Chapman and Hall; 1993..

Completing the process, we explored the sustainability of the instrument’s reduced versions, given that item anomalies could be revealed in the previously described CFA.

To increase the instrument’s efficiency, the items were dichotomized according to presence or absence of quality indicators (e.g., respectful treatment, adequate waiting time, and information on the procedure, among others). All the stages thus used probit models on tetrachoric matrices and the Weighted Least Square Mean and Variance adjusted robust estimator (WLSMV) ²⁶26. Muthén B, Asparouhov T. Latent variable analysis with categorical outcomes: multiple-group and growth modeling in Mplus, 2002. http://www.statmodel.com/download/webnotes/CatMGLong.pdf (acessado em 17/Set/2019).
http://www.statmodel.com/download/webnot... ^,2727. Finney SJ, DiStefano C. Non-normal and categorical data in structural equation modeling. In: Hancock GR, Mueller RO, editors. Structural equation modeling: a second course. Greenwich: Information Age Publishing; 2006. p. 269-314.. Three indices were used to assess the model fit ¹⁷17. Brown TA. Confirmatory factor analysis for applied research. New York: The Guilford Press; 2015.: Root Mean Square Error of Approximation (RMSEA), Comparative Fit Index (CFI), and Tucker-Lewis Index (TLI) ¹⁷17. Brown TA. Confirmatory factor analysis for applied research. New York: The Guilford Press; 2015.^,2222. Kline RB. Principles and practice of structural equation modeling. New York: The Guilford Press; 2015.^,2828. Hu L, Bentler P. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Struct Equ Modeling 1999; 6:1-55.^,2929. Hu LT, Bentler PM. Fit indices in covariance structure modeling: sensitivity to underparameterized model misspecification. Psychol Methods 1998; 3:424-53.^,3030. Tucker LR, Lewis C. A reliability coefficient for maximum likelihood factor analysis. Psychometrika 1973; 38:1-10.. RMSEA values below 0.60 suggest an admissible fit, while values above 0.10 indicate clear inadequacy and that the model should be rejected ²¹21. Muthén LK, Muthén BO. Mplus user's guide. 5th Ed. Los Angeles: Muthén & Muthén; 2007.^,3131. Browne MW, Cudeck R. Alternative ways of assessing model fit. In: Bollen KA, Long JS, editors. Testing structural equation models. London: Sage Publications; 1993. p. 136-62.. CFI and TLI vary from zero to one, and values above 0.95 indicate good fit ¹⁷17. Brown TA. Confirmatory factor analysis for applied research. New York: The Guilford Press; 2015.^,3030. Tucker LR, Lewis C. A reliability coefficient for maximum likelihood factor analysis. Psychometrika 1973; 38:1-10.. Interim diagnoses of cross-loadings (CFA) and residual correlations (ESEM and CFA) aimed at Modification Indices (MI) and the respective Expected Parameter Changes (EPC) offered in the outputs of the Mplus 8.1 software (https://www.statmodel.com/), used in the main analyses. The descriptive analyses used the Stata 15 software (https://www.stata.com).

Ethical aspects

The ethical principles of respect for the person, beneficence, and justice were ensured and have been described in a previous article ¹⁰10. Aquino EM, Menezes G, Barreto-de-Araújo TV, Alves MT, Alves SV, Almeida MdCCd, et al. Qualidade da atenção ao aborto no Sistema Único de Saúde do Nordeste brasileiro: o que dizem as mulheres? Ciênc Saúde Colet 2012; 17:1765-76.. The research protocol for GravSus-NE was approved by each respective Institutional Review Board and the Brazilian National Commission on Research Ethics.

Results

First stage

Of the 55 initially proposed items ¹⁵15. Aquino EML, Menezes GMS, Barreto-de-Araújo TV, Alves MT, Almeida MCC, Alves SV, et al. Avaliação da qualidade da atenção ao aborto: protótipo de questionário para usuárias de serviços de saúde. Cad Saúde Pública 2014; 30:2005-16., nine were excluded (D17 to D25), pertaining to the procedure’s timing. This was because 96% of the women had undergone curettage and could not answer the questions because they were under anesthesia. Nine other items on the context of care were removed because they contributed little to optimizing the instrument’s metric and scalar properties with items describing the type of procedure used in the uterine evacuation (A2); the type of professional performing the examination (A3, A4, A5, and P28) and/or furnishing information on contraceptives (P39); the type of persons present at the examination before or after the procedure (A12 and P31); or type of contraceptive prescribed (P42).

As the result of theoretical considerations (content), three derived indicators were established next, combining connected items: whether the woman received and understood information on health/physical condition (A7 and A8); whether there were persons present during the examination and whether this presence embarrassed the woman (A11 and A13); and whether the patient felt pain before the procedure, and if so, whether she was medicated (A14 and A15).

Some decisions were also made in light of the preliminary exploratory factor analyses of this set of items. Due to the high collinearity between three items pertaining to the post-procedure examination, two were excluded that qualified it - whether the treatment was respectful (P29) and whether privacy was ensured (P32), opting to maintain the examination itself (P28) as the marker of technical quality of care.

The item on supply of sanitary napkins (P47) moderately loaded the dimension on inputs/physical environment but showed an even greater connection with technical quality of care, consisting of items pertaining to pain management (P27) or blood pressure control (P50). Although the items might reflect a concern with the patient’s wellbeing and serve as an indicator of prevention of infection by avoiding the accumulation of material favorable to the development of microorganisms, these cross-loadings appeared difficult to interpret, thus justifying their removal.

The item on the presence of an accompanying person during hospitalization (P48) was excluded, since it did not load any dimension in the preliminary analysis and for substantive reasons. Although the item is not included in Brazil’s technical guidelines ¹²12. Secretaria de Atenção à Saúde, Ministério da Saúde. Atenção humanizada ao abortamento: norma técnica. Brasília: Ministério da Saúde; 2005. (Série A. Normas e Manuais Técnicos) (Série Direitos Sexuais e Direitos Reprodutivos - Caderno 4). since abortion is illegal and clandestine in Brazil, some women preferred to stay alone, thus preventing the item’s use as an indicator of quality of care.

Two items on reproductive planning - prescription of contraceptives (P40) and orientation on where to obtain the prescribed method (P43) - proved to be highly colinear, leading to estimation problems. We opted to maintain P43 because it showed greater reliability (expressed as a higher factor loading) and for a theoretical reason, since the orientation on where to obtain the method (P43) indicates completeness of this act, as opposed to mere prescription (P40), and thus better quality of care.

During this stage, the items’ dichotomization was backed by the results of the preliminary analyses, revealing maintenance of configural structure and even some improvement in the metric properties (e.g., increased reliability of the items). Details on the excluded items and respective alternative answers can be requested from the authors.

Second stage

At the end of the previous stage, there were 21 items that were then submitted to more detailed factor analyses. The initial four-factor proposal only fit moderately (Table 1). Although RMSEA revealed admissible fit values, the CFI and TLI values were borderline (0.95 > x > 0.90). All the factor correlations were moderate to low, the highest being 0.576, between F2 and F4 (not shown in the Table). Nevertheless, four of eight item loadings postulated a priori as belonging to F1 proved to be low (A1, A16, A11_13, and P54) in this factor. The loadings were also low for item A14_15 in F2 and item P46 in F3. By contiguity, the respective residuals (δ_i) were high, all above 0.80. Many modification indices suggested poor specification of the original configural structure, and that the latter merited more detailed scrutiny.

In light of this initial result, we focused on new exploratory analyses (ESEM with 2 to 6 factors). Although the preliminary analysis of the eigenvalues indicated the possible existence of up to six factors (eigenvalues > 1.0), this sixth factor lacked interpretability in the 6-factor ESEM, including only one item with moderate expression and with a cross-loading in F4 - λ_35(ƒ4) = 0.412 and λ_35(ƒ6) = 0.417. At the other extreme, the 2- and 3-factor ESEM showed even more cross-loadings (data not shown).

Table 2 shows the results of the four- and five-factor ESEM. The fits improved substantially, with all three indices at acceptable levels. Factor correlations remained relatively low in both models, the largest involving F2-F4 again in the four-factor model (𝜙 = 0.517) and F3-F4 in the five-factor model (𝜙 = 0.434). Due to the free estimation of cross-loadings in ESEM, the residuals were low in both models. No residual correlation was detected.

Table 2 clearly shows that four-factor Model 2 was unable to separate part of the manifests postulated in different dimensions, combining five items (A1, D16, A11_13, P54, and A6), purportedly related to the dimension reception and orientation, to three (P44, P45, and P46) connected to inputs/physical environment. On the contrary, there is an adequate separation in the five-factor model’s configural structure, which proved to be the most parsimonious and promising. Clearly the originally conjectured 4-factor model (Table 1) does not materialize in the ESEM, and it is even different from Model 2, which also has 4 factors.

Table 3 shows the CFA for the five-factor solution suggested in Model 3, further encompassing a cross-loading in A6 manifested consistently in the ESEM. The fit was similar to that of the related exploratory model. Three items (A11_13, A14_15, and P46) continued with low loadings, and complementarily with very high residuals. Again, no residual correlation was detected.

Aimed at dealing with these problematic items, Table 4 presents two alternative reduced models. Considering the values presented in Model 4 in Table 3, Model 5 excluded the items A6 (cross-loadings) and A14_15 (residual > 0.90); in Model 6, the exclusion extends to items A11_13 and P46, both with residuals > 0.8. The fit improved slightly in both solutions. Mainly in Model 6, all the loadings exceeded λ_i = 0.55, presenting residuals below 0.7 (the majority below 0.4). In this solution, the fourth factor has only two items. The interim diagnosis based on the MI/EPC does not show any cross-loading or presence of residual correlations.

Table 5 shows the square roots of the EMVs and the factor correlations concerning the last two models shown in Table 4. The conjectured DFV in the exploratory analyses appears to be sustained in the two alternative 5-factor models. All the ρvef• proved to be consistently higher than their respective correlations ϕf↔k. Focusing on F5 in Model 5, for example, ρvef5 widely outstrips the four correlations involving this factor (ϕ1↔5=0,11, ϕ2↔5=0,45,…ϕ3↔5=0,59 e ϕ4↔5=0,35). The same pattern extends to the other contrasts between the ρvef• and respective ϕf↔k, and all the differences were statistically significant (p < 0.001). Importantly, the FDV is exacerbated from Model 5 to Model 6, with ρvef4 increasing substantially from 0.61 to 0.71.

The final 17-item instrument and the respective options for answers are shown in an Annex.

Discussion

Quality of abortion care in Brazil has received relatively little attention ⁶6. Menezes G, Aquino EML. Pesquisa sobre o aborto no Brasil: avanços e desafios para o campo da saúde coletiva. Cad Saúde Pública 2009; 25 Suppl 2:S193-204.^,77. Departamento de Ciência e Tecnologia, Secretaria de Ciência, Tecnologia e Insumos Estratégicos, Ministério da Saúde. Aborto e saúde pública no Brasil: 20 anos. Brasília: Ministério da Saúde, 2009. (Série B. Textos Básicos de Saúde).. Our study intended to address this gap with a valid and efficient instrument in the Portuguese language for use in future studies.

The initially proposed four-dimensional conceptual model ¹⁵15. Aquino EML, Menezes GMS, Barreto-de-Araújo TV, Alves MT, Almeida MCC, Alves SV, et al. Avaliação da qualidade da atenção ao aborto: protótipo de questionário para usuárias de serviços de saúde. Cad Saúde Pública 2014; 30:2005-16. fit moderately, suggesting a distinct structure from the original. The subsequent analyses converged to some robust solutions. Model 6 was the most auspicious from the configural and metric point of view. The items have acceptable reliability, with all 17 factor loadings above 0.55, and 12 above 0.70. The residuals are within admissible margins (< 0.7), mostly below 0.4. This factor solution is consistent with the contents of the manifest items and respective dimensions; it shows the items’ factor specificity, expressed as the absence of apparent cross-loadings; it excludes item redundancy, manifested as the absence of striking residual correlations; and it displays FDV, as shown by the formal analysis contrasting the items’ aggregate informativity (by factor) and respective factor correlations.

This set of manifest items appears to adequately map five dimensions: reception, orientation, technical quality of care, inputs/physical environment, and continuity of care. As in the initial proposal, the latter three continue to be sustained in the new proposal, but the dimension reception and orientation partitions into two. The items postulated as manifestations of a purported cohesive set encompassing the sphere of contact at entry to the service form distinct dimensions, one involving reception and the other involving formation/orientation. Although both refer to health professionals’ communication with patients, they proved to have distinct contents.

Reception is defined by the Brazilian Ministry of Health as “decent and respectful treatment, listening, recognition, and acceptance of differences, respect for women’s and men’s right to decide, as well as access to care and case-resolution capacity” ¹²12. Secretaria de Atenção à Saúde, Ministério da Saúde. Atenção humanizada ao abortamento: norma técnica. Brasília: Ministério da Saúde; 2005. (Série A. Normas e Manuais Técnicos) (Série Direitos Sexuais e Direitos Reprodutivos - Caderno 4). (p. 17). In the QualiAborto-Pt questionnaire, the dimension reception consisted of three items: waiting for the first examination; waiting for the uterine evacuation procedure; and perception of discrimination during care.

The item on “respectful treatment in the first examination” (A6) presented cross-loading and was excluded. This may have occurred due to the term’s ambiguity in the Portuguese language. As intended in the original version in English, “respectful” can be the equivalent of courtesy and kindness; in Portuguese, the term also connotes decency (sometimes with a sexual content), which may have given rise to distinct interpretations of the item. Prior to the main psychometric analyses, an equivalent item (P29) pertaining to respectful treatment in the post-procedure examination had been eliminated due to the high degree of collinearity with privacy (P30_32) and the examination itself (P28). In the instrument’s refinement, the latter was maintained, since it was more objective for assessing the dimension technical quality of care, discussed below. The item on privacy was maintained, conceived as preservation of intimacy in the body’s exposure and manipulation ³²32. Anjos MF, Fortes PAC. Desafios para a preservação da privacidade no contexto da saúde. Bioética 2000; 8:307-22.^,3333. Pupulim JSL, Sawada NO. Privacidade física referente à exposição e manipulação corporal: percepção de pacientes hospitalizados. Texto & Contexto Enferm 2010; 19:36.. The literature defines it as an expression of decent treatment and respectful care ³⁴34. Vaitsman J, Andrade GRB. Satisfação e responsividade: formas de medir a qualidade e a humanização da assistência à saúde. Ciênc Saúde Colet 2005; 10:599-613.. We thus recommend revisiting these items in the future.

Also prior to the main analyses, another item that presented problems and was removed concerned the presence of an accompanying person during part or all of the hospital stay (P48). Although not present in the Brazilian guidelines, its beneficial effect on childbirth care ³⁵35. Diniz CSG, d'Orsi E, Domingues RMSM, Torres JA, Dias MAB, Schneck CA, et al. Implementação da presença de acompanhantes durante a internação para o parto: dados da pesquisa nacional Nascer no Brasil. Cad Saúde Pública 2014 Suppl 1; 30:S140-53. and recently on abortion care ³⁶36. Altshuler AL, Nguyen BT, Riley HE, Tinsley ML, Tuncalp O. Male partners' involvement in abortion care: a mixed-methods systematic review. Perspect Sex Reprod Health 2016; 48:209-19. is acknowledged, emphasizing the pertinence of revisiting this indicator in future analyses. However, since abortion is illegal and clandestine in Brazilian, some women prefer to stay alone at the hospital, which relativizes the item’s importance as an appropriate manifestation of quality of care.

The other items in this dimension displayed good psychometric properties. Two items measure the adequacy of waiting time until the first examination (A1) and until the procedure (A16), where the care’s speed and timeliness are considered essential to the outcome in abortion complications ⁵5. Santana DS, Cecatti JG, Parpinelli MA, Haddad SM, Costa ML, Sousa MH, et al. Severe maternal morbidity due to abortion prospectively identified in a surveillance network in Brazil. Int J Gynaecol Obstet 2012; 119:44-8..

Likewise, the item on perception of discrimination in care (P54), which aims to grasp respect for differences, showed good psychometric properties and remained in the final model. This item is especially relevant, since in the hierarchy of priorities of care, beyond technical criteria, other (subjective and moral) items intervene, lending primacy to childbirth rather than to abortion complications ³⁷37. McCallum C, Menezes G, Reis AP. O dilema de uma prática: experiências de aborto em uma maternidade pública de Salvador, Bahia. Hist Ciênc Saúde-Manguinhos 2016; 23:37-56..

The partitioned dimension - orientation - assumes “the transfer of necessary information for conducting the process of care with the woman as protagonist in the health act, decision-making, and selfcare in keeping with the SUS guidelines” ¹²12. Secretaria de Atenção à Saúde, Ministério da Saúde. Atenção humanizada ao abortamento: norma técnica. Brasília: Ministério da Saúde; 2005. (Série A. Normas e Manuais Técnicos) (Série Direitos Sexuais e Direitos Reprodutivos - Caderno 4). (p. 17). This dimension was addressed properly and included, in the first place, an item that assesses the transfer of information on the woman’s physical condition and whether she has understood the information (A7_8). Another item (A9) considers the transfer of information on what will happen during the procedure. Both concern explanations on health conditions, considered essential for ensuring the rights to information and autonomy ³⁴34. Vaitsman J, Andrade GRB. Satisfação e responsividade: formas de medir a qualidade e a humanização da assistência à saúde. Ciênc Saúde Colet 2005; 10:599-613.. The item on the opportunity to ask questions (A10) reflects the health team’s capacity to hear and answer patients’ questions prior to the procedure, considered a crucial component in quality of care and an indicator of two-way communication ³⁸38. Nascimento MI, Reichenheim ME, Monteiro GTR. Estrutura dimensional da versão brasileira da Escala de Satisfação com o Processo Interpessoal de Cuidados Médicos Gerais. Cad Saúde Pública 2011; 27:2351-63..

The items on technical quality of care showed good psychometric properties, and four out of five postulates remained in the final model. These address factual and objective information ³⁹39. McDowell I. Measuring health: a guide to rating scales and questionnaires. Oxford: Oxford University Press; 2006. such as performing low-density technologies and universally disseminated knowledge - measurement of body temperature (P50) and blood pressure (P51) or control of bleeding (P52). Appreciation of pain management was the only item with less-than-admissible performance, possibly due to the greater subjectivity in the capacity to withstand pain ³⁹39. McDowell I. Measuring health: a guide to rating scales and questionnaires. Oxford: Oxford University Press; 2006.. This aspect merits more in-depth examination in the future, given its importance for quality of care from the perspective of reproductive rights, the right to health, and physical integrity.

In the dimension inputs/physical environment, two criteria related to the environment that remained in the final instrument - cleanliness of the physical space and changes to bedclothing -, are traditionally included in questionnaires ³⁴34. Vaitsman J, Andrade GRB. Satisfação e responsividade: formas de medir a qualidade e a humanização da assistência à saúde. Ciênc Saúde Colet 2005; 10:599-613. assessing women’s healthcare perceptions (e.g., the World Health Survey⁴⁰40. World Health Organization. World Health Survey, 2012. https://www.who.int/healthinfo/survey/en/ (acessado em 09/Jun/2012).
https://www.who.int/healthinfo/survey/en... , from which they were extracted). These indicators aim to grasp aspects connected to the services’ infrastructure and facilities, providing information on the available resources for provision of care.

Two other items proposed by the research team - supply of patient’s hospital clothing with the correct size and sanitary napkins - were not confirmed as good manifestations and were eliminated for different reasons. The supply of sanitary napkins (P47) presented a cross-loading in another factor, starting in the preliminary analyses. Signaling heterogeneity in the patients’ perceptions, some saw this item as related to technical quality of care while others as an input related to the quality of the physical environment where the care is provided. The other item (P46) concerning the supply of patient’s clothing was eliminated in the principal stage of analysis with 21 items, due to its low reliability. It was related to the size of the clothing - aimed at measuring privacy and non-exposure of the body - and not to the supply per se or to the clothing’s cleanliness ³⁴34. Vaitsman J, Andrade GRB. Satisfação e responsividade: formas de medir a qualidade e a humanização da assistência à saúde. Ciênc Saúde Colet 2005; 10:599-613.. Both items merit future scrutiny, since one cannot rule out the possibility of problems in drafting these questions, developed more from a descriptive perspective than to compose a scale. It would also be relevant to identify other items, such as quality of meals, airiness and silence in the environment, and comfort in the facilities ³⁴34. Vaitsman J, Andrade GRB. Satisfação e responsividade: formas de medir a qualidade e a humanização da assistência à saúde. Ciênc Saúde Colet 2005; 10:599-613., since this dimension (unlike the others) now has only two items, which is undesirable for good dimensional mapping ⁴¹41. Wilson M. Constructing measures: an item response theory approach. Mahwah: Lawrence Erlbaum; 2005..

The dimension continuity of care includes relational aspects in communication between healthcare providers and patients. The items on orientation for post-discharge care (P33), information on family planning (P36), orientation on where to obtain contraceptive methods (P43), and orientation on risk of a new pregnancy (P34) displayed good psychometric properties and remained in the model. When included in this dimension together with the item on scheduling the follow-up appointment (P35), they corroborate the literature on their pertinence to continuity of care beyond the moment assessed ³⁴34. Vaitsman J, Andrade GRB. Satisfação e responsividade: formas de medir a qualidade e a humanização da assistência à saúde. Ciênc Saúde Colet 2005; 10:599-613.^,4242. Tumlinson K. Measuring quality of care: a review of previously used methodologies and indicators. New York: Population Council; 2016..

As a continuation of the work launched with the elaboration of a prototype to assess the quality of abortion care, the current evaluation of psychometric properties in the QualiAborto-Pt questionnaire was based on the concept’s definition and its components, the cross-cultural adaptation of a set of items proposed by the WHO, and the adaptation and formulation of other items that expressed the criteria defined in the conceptual model.

Items not included in the current analyses can be incorporated in the future and should be tested in new analyses. The type of method used in the uterine evacuation would be the first of these, since Brazilian and international guidelines recommend the use of manual or electric aspiration and medical (drug-induced) abortion ¹²12. Secretaria de Atenção à Saúde, Ministério da Saúde. Atenção humanizada ao abortamento: norma técnica. Brasília: Ministério da Saúde; 2005. (Série A. Normas e Manuais Técnicos) (Série Direitos Sexuais e Direitos Reprodutivos - Caderno 4).^,1414. World Health Organization. Maternal health and safe motherhood programme. Studying unsafe abortion: a practical guide. Geneva: World Health Organization, 1996.. Both involve less risk and greater patient safety than curettage, which is still widely used in Brazilian hospitals. Despite its pertinence, the item’s inclusion depends on the women’s knowledge of the methods in order for them to respond “reliably”.

A second potential indicator relates to ultrasound performed in the hospital and the waiting time for performing this test. Ultrasound has become an essential tool for diagnosing the type of abortion and determination of the therapeutic approach. The lack of ultrasound on the hospital night shift and weekends leads to delays in care and longer length of hospital stay ⁴³43. Lima MRP. Práticas e significados em torno da ultrassonografia obstétrica e aborto em Salvador-Brasil [Dissertação de Mestrado]. Salvador: Universidade Federal da Bahia; 2016..

A third indicator to be examined in the future is sharing the same space with postpartum women and their newborn infants. Studies have shown that this embarrasses women undergoing abortion, especially during visiting hours, when they are asked where their babies are ⁴⁴44. Carneiro MF, Iriart JAB, Menezes GMS. "Largada sozinha, mas tudo bem": paradoxos da experiência de mulheres na hospitalização por abortamento provocado em Salvador, Bahia, Brasil. Interface (Botucatu, Online) 2013; 17:405-18..

A contingency in the current study involves the exclusion of nine items on the uterine evacuation’s timing, since 96% of the women underwent curettage and were unable to assess the care during the procedure because they were under anesthesia. Future studies can elucidate the properties of these items, related strictly to the type of uterine evacuation procedure.

Another question that requires greater exploration concern the items that were combined in the current study and assumed as single questions, but which were obtained on the basis of separate questions. Combining information in data’s processing and analysis is not necessarily equivalent to unifying their contents. Future studies can clarify this question, because the semantic union allows decreasing the time spent applying the questionnaire and increasing its efficiency.

A potential limitation to the study is the fact that the models were tested, modified, and corroborated (some) in a single data set. Ideally, the findings’ corroboration (or lack thereof) would be done in new studies with similar or even different domains, also aimed at assessing configural, metric, and scalar invariance ¹⁷17. Brown TA. Confirmatory factor analysis for applied research. New York: The Guilford Press; 2015..

In fact, the unexplored scalar structure is another drawbasck to mention, but this stems less from a flaw in the article itself than from the instrument’s own ongoing development process. Beyond good evidence on configural and metric structure, it is necessary to affirm an instrument’s mapping capacity in order to endorse it definitively ⁴¹41. Wilson M. Constructing measures: an item response theory approach. Mahwah: Lawrence Erlbaum; 2005.^,4545. Reichenheim ME, Hökerberg YHM, Moraes CL. Assessing construct structural validity of epidemiological measurement tools: a seven-step roadmap. Cad Saúde Pública 2014; 30:927-39.. Future studies might assess the component items’ scalability and the scales formed by them ⁴¹41. Wilson M. Constructing measures: an item response theory approach. Mahwah: Lawrence Erlbaum; 2005.^,4646. Sijtsma K, Molenaar IW. Introduction to nonparametric item response theory. v. 5. Thousand Oaks: Sage; 2002. (Measurement Methods for the Social Science)..

The questionnaire’s application during the patient’s hospital stay was intended to avoid difficulties in obtaining interviews in surveys on abortion, aggravated by the procedure’s illegality in Brazil ⁴⁷47. Barreto T, Campbell OM, Davies JL, Fauveau V, Filippi VG, Graham WJ, et al. Investigating induced abortion in developing countries: methods and problems. Stud Fam Plann 1992; 23:159-70.. Strategies to hide the practice include omission or denial of abortion in household interviews, even when there are prior records of hospitalization for this cause, as well as reporting of imprecise contact information by the patient during hospitalization ⁴⁸48. Osis MJD, Hardy E, Faúndes A, Rodrigues T. Dificuldades para obter informações da população de mulheres sobre aborto ilegal. Rev Saúde Pública 1996; 30:444-51.. Follow-up studies outside the hospital context have reported losses of up to 60% ⁴⁷47. Barreto T, Campbell OM, Davies JL, Fauveau V, Filippi VG, Graham WJ, et al. Investigating induced abortion in developing countries: methods and problems. Stud Fam Plann 1992; 23:159-70.^,4949. Söderberg H, Andersson C, Janzon L, Sjöberg N-O. Selection bias in a study on how women experienced induced abortion. Eur J Obstet Gynecol Reprod Biol 1998; 77:67-70.^,5050. Bradshaw Z, Slade P. The effects of induced abortion on emotional experiences and relationships: a critical review of the literature. Clin Psychol Rev 2003; 23:929-58..

The production of quality-of-care indicators based on the woman’s report at the time of hospital discharge reduces the potential selection bias but requires considering the so-called “gratitude bias” ³⁴34. Vaitsman J, Andrade GRB. Satisfação e responsividade: formas de medir a qualidade e a humanização da assistência à saúde. Ciênc Saúde Colet 2005; 10:599-613.. The latter is heavily dependent on recall conditions, especially when a complex procedure with major emotional burden has just been performed. Future studies should consider other strategies to assess the degree to which the current psychometric results are actually influenced by this problem.

In short, the results indicate that the current 17-item version of QualiAborto-Pt can already be recommended for use in Brazil to assess the quality of care for unsafe abortion, despite persistent questions for further study. With adjustments and adaptations, its use can be extended to other sociocultural contexts, including other Portuguese-speaking countries and those with restrictive abortion laws. Its use should be encouraged, not only for substantive purposes in evaluative studies on care for unsafe abortion, but to generate backing for future studies on its internal structure and external validity ⁵¹51. Mokkink LB, Terwee CB, Knol DL, Stratford PW, Alonso J, Patrick DL, et al. The COSMIN checklist for evaluating the methodological quality of studies on measurement properties: a clarification of its content. BMC Med Res Methodol 2010; 10:22.. The refinement of high-quality instruments such as the QualiAborto-Pt questionnaire can contribute to the comparability of studies and thus to better quality of care.

Acknowledgements

The authors wish to acknowledge MCT/CNPq/MS-SCTIE-DECIT/CT, MS/CNPq/FAPESB/SESAB/PPSUS, FAPEMA/SES, MA/MS/CNPq/PPSUS, and CNPq, as well as Eleonora Schiavo, Lilian Marinho, and Liberata Coimbra, who contributed to the GravSus-NE research project’s conception and development, and the 2,804 women users of the Unified Health System (SUS), who generously granted interviews for this study. In memoriam: Luci Praciano Lima, a dear partner in this study.

References

Publication Dates

History