2.1 Study design and sampling
A quasi-experimental, non-randomized, controlled study with pre- and post-test assessments was conducted among ASHAs of two selected districts. In each of the two selected districts, two control and intervention blocks were selected. The selection of intervention blocks was carried out randomly; simultaneously, control blocks were selected adjacent to the two intervention blocks in the respective districts. As this was a panel study, the same sampled respondents were provided training, and the same individuals were interviewed during the baseline and end line rounds. This allowed assessment of the trajectories of individual outcomes in the intervention group and compared with those of the control group.
2.2 Sample Size and participation
N= 2(Zα + Zb)2 × p(1-p)
(P1- P0)2
Where N = Sample size required for each group
P1= Probability of event in the intervention group = 75%
P0= Probability of event in the control group = 50%
p= (P1 + P0)/2
Zα = Standard normal deviate corresponding to the level of significance (type I error rate)
Zb = Standard normal deviate corresponding to the chance of not detecting the relative risk as significant (type II error rate)
A sample size of 60 per group was derived assuming that 50% of ASHAs had knowledge on soft skills during the baseline and hypothesized that this would increase to 75% after the intervention at 95% confidence level and power of study as 80%. The sample size was equally divided in each intervention or control block. Within each block, respondents were selected randomly for the interview.
2.3 Data collection
Data were collected by ten trained investigators. Structured questionnaires were used to collect data in both the survey rounds. The questionnaire had the following sections: a) socio-demographic characteristics, b) knowledge and practices related to soft skills (communication skills, self-confidence, problem–solving and decision-making skills, time and stress management skills), and financial literacy. The socio-demographic characteristics included age, years of schooling, years of experience of working as ASHA, monthly income from the job, and social class ASHA belongs to. The four domains of soft skills and financial literacy were assessed through different scales. The communication skills scale consisted of 7 questions with scores ranging between 0-17. The self-confidence scale had 6 questions on a five-point Likert scale (no confidence to very confident) with scores ranging between 6 and 30. There were 4 questions in the problem-solving and decision-making scale, and the minimum and maximum scores were 0 and 11, respectively. Similarly, the time and stress management scale had 5 questions, and the scores ranged between 0-20. Assessment of financial literacy was based on 5 questions with minimum and maximum scores of the scale ranging between 0 and 24, respectively. The scores of the individual question in the scale of each domain were summed up to calculate their aggregate scores. All the scales were validated, and the reliability scores (Cronbach’s alpha score) of all the scales were found more than 0.7.
Additionally, during the post-test assessment, current practices and related perceptions of ASHAs from the intervention groups were also captured. Here, we specifically assessed general views on P.A.C.E. training module and perceptions regarding the change in practices related to gender issues, soft skills (problem-solving, communication, and time management), and savings at the personal level, community level, and workplace. Separate scales for assessing each of these three domains (changes in perceptions at the personal level, community level, and workplace) were developed. The questions in all the scales were based on a four-point Likert-scale. The responses varied from completely agree to completely disagree. The scores of the questions on every scale were aggregated with completely agree given a score of 2 and agree a score of 1 and rest 0. The maximum and minimum scores of changes at the personal level varied between 0 and 22, community-level between 0 and 18, and the workplace between 0 and 14. All the scales were validated, and the reliability scores (Cronbach’s alpha score) of all the scales were observed more than 0.85. For assessing general perceptions on P.A.C.E. training modules, the participants were asked to rate the quality of the training on a scale from 1 to 10.
The questionnaires were standardized, translated into the local language (Hindi), and field-tested before data collection. We used Computer Assisted Personal Interview (CAPI) for collecting quality real-time data during both the survey rounds. To collect quality of data, two supervisors, one in each district, were assigned to randomly back-check and spot-check 10% of all interviews during both the rounds.
2.4 Data analysis
The data were analyzed using the IBM SPSS Statistics for Windows version 24.0 (IBM Corp., Armonk, N.Y., USA). Descriptive data were expressed as frequency or percentages for categorical variables and mean (Standard Deviation, SD) or median (Interquartile Range, IQR) for continuous variables. The paired t-test (or Wilcoxon signed-rank test for medians) was conducted to assess the differences between average pre-and post-test scores in the intervention and control groups for four domains of soft skills and financial literacy. Factor analysis was used to derive change patterns of the training. We used this method to assess the degree to which the four domains of soft skills, financial literacy, and domains of perceptions of change at the personal level, community level, and workplace in the intervention group were correlated with each other and to derive a new set of composite variables. These new set of composite variables, not related to each other, represent discrete change patterns of the training. Only the change patterns with eigenvalues >1.0 were included in the analysis. The domains that loaded highly (|>0.30|) in varimax rotated change patterns were shown in the analysis. The Kaiser-Meyer-Olkin (KMO) measure reached the acceptable limit of 0.6, and Bartlett’s test of sphericity was significant (p<0.001), meaning thereby that the data were suitable for factor analysis.
A general linear model of regression was performed to assess the association between the change pattern scores and socio-demographic variables using main-effect analysis. Standard regression coefficients (β) and 95% confidence intervals were used to depict the strength of and precision of associations A two-sided p-value <0.05 was considered statistically significant.
2.5 Ethical considerations
The study was granted ethical approval by MAMTA Ethical Review Board. Informed consent was obtained from all the study participants.