Study design, setting and period
A retrospective single cohort study design was conducted in Debre Markos comprehensive specialized Hospital (DMCSH), which is located in Derbre Markos Town, Northwest, Ethiopia. It was conducted among 302 mothers diagnosed with postpartum hemorrhage from January 1, 2016 until December 31, 2020. Derbre Markos Town is located about 299 km far from Addis Ababa, the capital city of Ethiopia and 265 km from Bihar Dar. DMCSH is providing service for more than 5 million people per year. In the maternity department, a total of 1 emergency surgeon 1MSc clinical midwife specialist, 41 general practitioners, 16 obstetrics and gynecology specialists and 44 midwives. The department has 60 beds for inpatient clients to serve high-risk mothers, gynecologic patients, and postnatal mothers. The annual delivery report showed that, 6017 mothers gave birth in the hospital.
Population
All women diagnosed with postpartum hemorrhage in Debre Markos comprehensive specialized hospital, during the period of Jan.1, 2016 to Dec. 31, 2020 G.C were taken as source and study populations. Clear record of diagnosis/PPH/ was used as an inclusion criteria. The diagnosis might be “anemia secondary to postpartum hemorrhage” or “hypovolemic shock/shock secondary to postpartum hemorrhage”. These diagnoses were included. Unclear anemia/shocks were excluded.
Sample Size Determination
The sample size was calculated by using a sample size calculation for survival analysis in STATA 14. Sample size (n) = number of event / Probability of event.
$$\mathrm{Number of Event }=\frac{(\mathrm{Z a}/2+\mathrm{ ZB})2 }{\mathrm{pq }(\mathrm{logHR})2}$$
By taking the hazard ratio of Volume of IV fluids given in the first hour with NASG group (1.575) from a study conducted on time to recovery from shock for women with obstetrics hemorrhage in Egypt [2], the sample size was calculated. The assumptions for this were, probability of withdrawals (0.1), probability of developing an event (0.5), alpha (0.05), and power of (80%). Using STATA command (Stpowerlogrank 0.5, hratio (1.575) power (0.8) wdprob (0.1)), it becomes 302.
Sampling technique
By using registration book in maternity ward of DMCSH, maternal charts were retrieved and then data on individuals with postpartum hemorrhage was gathered starting from first date of diagnosis record up to the last record of the same diagnosis in the same period. All mothers with postpartum hemorrhage in the years 2016 to 2020 who fulfill inclusion criteria were consecutively included in the study. The total postpartum mothers in the hospital registered as anemia and or hypovolemic shock within the five years period were 375 and of these 328 fulfills the eligibility criteria. From these again 302 were selected consecutively because the total population was small to apply other probability sampling methods (Fig. 1).
The Study variable
Dependent variable
Time to recovery from postpartum hemorrhage (in hours) (i.e., the event of interest is recovery).
Independent variable
Socio-demographic
Age at diagnosis, marital status, residence, place of delivery.
Treatment-related variable
Fluid resuscitation, Blood transfusion, AMTSL (Active Management of Third Stage of Labor), NASG utilization, tranexamic acid utilization, MRP, Antibiotic used, types of surgery done.
Obstetrics related factors
Mode of delivery, complication of labor, Episiotomy, Birth weight, Gravidity, Parity, multiple pregnancy, antepartum iron intake, number of visits for ANC, status of anemia during the antepartum period, types PPH.
Comorbidities
HIV, TB, Syphilis, Hypertension, Hepatitis.
Operational definition
Recovered
Women were considered as recovered when no more vital sign derangement, which means pulse rate < 100, and SBP ≥ 100 [13] and hematocrit greater than 21% after diagnosed for postpartum hemorrhage.
Start time
The time of diagnosis of PPH.
Follow-up period
Time of diagnosis to the first record of stable vital sign and/or hematocrit > 21%
Time to recovery
Was obtained by calculating the difference (in hours) from time of diagnosis until the last recorded deranged vital sign and/or last record of hematocrit < = 21%
Event
The occurrence of stable vital sign and/or hematocrit > 21% after the diagnosis of postpartum hemorrhage.
Incomplete chart
Cards with no baseline data about diagnosis of PPH and when variables that affect the outcome variable in the study were missed such as, vital sign (pulse and blood pressure) and date of diagnosis.
Censored
Declared when the outcome of interest was not observed, like cases transferred to other facility.
Postpartum hemorrhage
Considered as PPH case if the health care provider diagnosed and recorded on the chart as PPH [14].
Co-morbidity
Is the presence of one or more diseases other than postpartum hemorrhage. The presence of any of these conditions during the follow-up was designated a “Yes‟, while the absence of these conditions was denoted as “No‟. Those comorbidities were accessed from the patient’s card.
Data collection tool
A structured data abstraction checklist was adapted from different literatures and standard treatment protocols for PPH.
Data collection procedure
The information was collected from the operation notes, admission and discharge delivery registration logbook, and patient cards. Data were extracted for one month by using a data extraction sheet that was prepared in the English language after reviewing different literatures and from the standard treatment for the management of PPH. Two midwives working in the maternity ward retrieved patients’ charts and extracted the required data under supervision.
Data quality assurances
To ensure the quality of data at the beginning, a data collection tool was pre-tested on 5% of the calculated sample size at Injibara General Hospital, which is found about 200 km from the actual study setting. Then, necessary changes were made based on the gaps identified. The abstraction format was checked to the hospital documentation system to make sure that the agreement of the data abstraction format with the need for the study. Any error found during the process was corrected and changes were made to the last version of the data abstraction format. Training on record review was given to the data collectors and supervisors for one day before the actual data collection. All collected data were reviewed for completeness and consistency during data cleaning and analysis.
Data processing and analysis
Data were entered into Epi-Data version 4.2 after checking for completeness, and then the data were cleaned and transferred to STATA 14 statistical software for analysis. Schoenfeld residual analysis (global test) was used to show the cox proportional hazard model assumption validity. Predictors of time to recovery were identified using bivariable and multivariable cox proportional hazard models (CPHM). All independent variables that had a P value less than 0.25 in the bivariable analysis were considered as candidate variables for the multivariable analysis. The output of the multivariable CPHM was presented using adjusted hazard ratios with the respective 95% confidence intervals (CI) to measure the strength of association. Kaplan–Meier survival estimator was used to estimate median recovery time, while log-rank tests to compare time to recovery between groups. Before running the cox regression model, assumption of proportional- hazards and multi-collinearity were checked. The Cox Snell residual was used to assess model fitness and proportional hazard assumption.
Ethical Considerations
An ethical approval letter was obtained from the Institutional Research Ethics Review Committee of college of health science, Debre Markos University. Following the approval, an official letter of cooperation was written to the concerned bodies by the college to facilitate the support and commitment of responsible bodies. Permission was obtained from the hospital medical director and department of gynecology and obstetrics/maternity. As the study was conducted through review of chart records, the individual patients were not subjected to any harm and personal identifiers during data collection and confidentiality was maintained through not recording the name of the participant on the checklist.