Assessing completeness of patient medical records of surgical and obstetric patients in Northern Tanzania

ABSTRACT Background Strengthening surgical services in resource-constrained settings is contingent on using high-quality data to inform decision making at clinical, facility, and policy levels. However, the evidence is sparse on gaps in paper-based medical record quality for surgical and obstetric patients in low-resource settings. Objective We aim to examine surgical and obstetric patient medical record data quality in health facilities as part of a surgical system strengthening initiative in northern Tanzania. Methods To measure the incidence of Surgical Site Infections (SSIs), sepsis and maternal sepsis surgical and obstetric inpatients were followed prospectively, over three months in ten primary, district, and regional health facilities in northern Tanzania. Between April 22nd to May 1st, 2018, we retrospectively reviewed paper-based medical records of surgical and obstetric patients diagnosed with SSIs, post-operative sepsis, and maternal sepsis in the three-month follow-up period. A data quality assessment tool with18 data elements related to documentation of SSIs and sepsis diagnosis, their respective symptoms and vital signs, inpatient daily monitoring indicators, and demographic information was developed and used to assess the completeness of patient medical records. Results Among the 157 patients diagnosed with SSI and sepsis, we found and reviewed 68% of all medical records. Among records reviewed, approximately one third (34%) and one quarter (23%) included documentation of SSI and sepsis diagnoses, respectively. 6% of reviewed records included documentation of all SSI and sepsis diagnoses, symptoms and vital signs, inpatient daily monitoring indicators, and demographic data. Conclusions Strengthening data quality and record-keeping is essential for surgical team communication, continuity of care, and patient safety, especially in low resource settings where paper-based records are the primary means of data collection. High-quality primary health information provides facilities with actionable data for improving surgical and obstetric care quality at the facility level.


Background
The Lancet Commission has primarily driven the growing global movement to expand access to essential and emergency surgical and anesthesia care on Global Surgery report in 2015, which identified a gap of 143 million surgical procedures in low-income and middle-income countries (LMICs) to meet the need [1]. This movement, along with others, has brought the importance of surgical quality to the forefront of global health priorities [2,3]. To address the high burden of conditions amenable to surgical care in Tanzania, the Ministry of Health, Community Development, Gender, Elderly, and Children launched its first National Surgical, Obstetric, and Anesthesia Plan in 2018 [4,5]. However, the evidence on improving surgical quality and generating scientific evidence through high-quality data is sparse. Such improvement is contingent upon the availability of reliable data to assess the current status of surgical care quality, evaluate the need, identify the gaps and thus inform priorities both at the facility and national policy levels [6,7]. Effective health quality improvement depends on the complete, accurate, and timely flow of data between primary, secondary, and regional health facilities and the central data repository [8]. At the facility level, having access to reliable and complete patient data is essential for surgical and obstetric providers as well as hospital managers to coordinate care, identify areas for improvement and ensure effective and efficient delivery of safe, high-quality care [9]. While there has been an increase in the availability and utilization of electronic medical records in sub-Saharan Africa over the last decade, there are significant gaps in the infrastructure to support the use of electronic records in LMICs [10]. Since paper-based record-keeping is still the primary medical data collection method in LMICs, the existing paper records must be better utilized to ensure patient safety and high-quality surgical and obstetric patient care.
Tanzania has a two and a half decade history of using a Health Management Information System (HMIS), also known as Arrifa Aa Uendeshji Huduma za Afya (MTUHA), a paper-based data entry system of 12 HMIS booklets [11]. These booklets are meant to cover all health services and are mandatory to be completed and reported to the district health authority through a national system. Their focus is primarily on maternal and child health indicators. Surgical care indicators, such as rates of surgical site infections (SSIs), post-operative sepsis, and maternal sepsis, which are often used as measures of surgical quality, are not routinely reported [11,12]. Due to the absence of mandatory national reporting for these indicators, patient medical data are often poorly recorded and underutilized, which creates barriers for peri-operative monitoring, continuity of care, and patient safety [11,13]. Medical records for surgical patients, which are separate from the MTUHA system, help bridge this gap. Even though medical records for surgical and obstetric patients vary by facility, they typically include essential elements such as patient history, a partograph for obstetric patients, post-operative notes, and a drug administration chart. Medical records are used as the primary means of documentation of perioperative patient clinical information, and communication between the surgical and ward teams. The importance of medical records in the continuity of perioperative patient care warrants a thorough exploration of the current gaps in data recording to identify areas of improvement, although the authors found no published literature on the state of surgical data collection in Tanzania To better understand the surgical data quality, we examined the completeness of surgical and obstetric patient medical records in 10 health facilities in northern Tanzania, focusing on the essential components needed for perioperative patient monitoring. To our knowledge, this is the first quantitative assessment of medical record completeness in primary, district, and regional facilities in Tanzania.

Identification of SSI and sepsis patients
Our data quality assessment was conducted as part of a longitudinal, multi-centered, quasi-experimental study to examine the impact of a multi-component intervention in ten intervention and ten control facilities in Tanzania's Lake Zone region. The study protocol is published elsewhere [14]. Briefly, the study included 4 health centers, 11 district hospitals, and 5 regional referral hospitals. On average, 82 surgical procedures were conducted monthly per facility, which was mostly cesarean sections.
The baseline study was conducted between February 1st, 2018, and May 1st, 2018, by recently graduated Tanzanian medical doctors who had one or two years of specialized training in clinical and/or surgical care and were trained as data collectors. The data collectors followed surgical inpatients prospectively using a screening tool to diagnose SSI, postoperative sepsis, and maternal sepsis for the duration of the patient's stay up to 30 days. All patients who were over the age of 5, had undergone surgery in a major operating room or delivered at the facility vaginally or through a cesarean section were included in the study. All minor surgeries, visiting surgeons' patients, antenatal patients, women with spontaneous abortions, and surgical outpatients were excluded. The list of minor surgeries excluded from the study is listed in Appendix I. To make the diagnosis, data collectors conducted daily patient rounds independently of the surgical teams and reviewed patient charts. When information could not be obtained through clinical examination or chart review, data collectors discussed cases with surgical providers.
The research team developed the SSI screening tool based on the CDC criteria for diagnosing and classifying SSI [15]. Both the post-operative and maternal sepsis screening tools were adapted for lowresource settings from Surviving Sepsis Campaign guidelines, which was based on Second International Consensus on Sepsis [16][17][18]. The tools were pretested by the research team and the data collectors at two secondary health facilities in Dar es Salaam. To ensure consistency in interpretation, the research team and data collectors debriefed the process and identified inconsistencies, unclear questions, or questions that were not relevant to the local context and made the appropriate adjustments.

Study design and data collection
From April 22nd, 2018, to May 1st, 2018, we retrospectively reviewed the medical records of surgical and obstetric patients who were diagnosed with SSI and sepsis during the three-month baseline period. During the two weeks, data collectors searched independently for patient files in the medical records departments and surgical wards, and then a member of the research team and a data collector from each of the facilities searched together on a designated chart review day. At this point, if the patient files were not found, then it was considered 'not found.' The level completeness of the medical records was assessed by evaluating the rates of 1) documentation of SSI or sepsis and their respective symptoms and vitals, 2) patient daily monitoring and progress information, and 3) documentation of key patient demographics. A 'complete' medical record included all of these three components. To operationalize documentation of SSI, we defined SSI as 'pus draining from the wound,' 'closed wound opened,' 'wound with a foul smell,' and 'wound infection [19].' Since postoperative complication documentation mechanisms are not uniform across region or health facility, we developed separate criteria to assess documentation of these three components in medical records and, thus, completeness of a medical record (Table 1).
To examine the level of completion of medical records, we developed a Data Quality Assessment (DQA) tool which included 18 questions addressing three main areas: 1) documentation of SSI, sepsis and their respective symptoms and vitals (Questions: 14,15,16), 2) patient daily monitoring and progress (Questions: 5, 6,7, 11, 12), and 3) patient characteristics (Questions: 3, 4). (Appendix II). The tool was developed by the research team and was informed by the World Health Organization (WHO) Western Pacific Region's Medical Records Manual: A Guide for Developing Countries [20]. Development of the tool was also informed by 5 or 6 semi-structured interviews with key stakeholders (i.e. surgical providers, chief medical officers, surgical nurses, and data managers) by members of the research team to understand data flow from the facility to the national level.

Data analysis
Data for all found medical records were collected in Excel (Microsoft, Redmond, WA, USA) and were analyzed in Stata/IC 15.1 (College Station, TX, USA). Descriptive analyses were conducted to describe the demographic characteristics of the study population and to summarize medical record completeness rates.

Results
Demographics: During the three-month baseline period, 7%, 6%, and 2% of all post-operative and post-partum patients followed in the 10 intervention facilities (N = 4,343) developed an SSI, postoperative sepsis and maternal sepsis respectively. The procedure most frequently leading to a complication was a cesarean section, with 69% and 53% leading to an SSI and sepsis, respectively (Table A1/A2). A little over one-third of all patients diagnosed with an SSI were in-hospital for over 10 days (34%), while this rate was slightly lower for those diagnosed with sepsis (31%) Table 1A/2A. SSIs and sepsis were diagnosed 6.4 and 4.0 days after hospitalization on average, respectively. Most cases of SSI were superficial (70%) (Appendix III). Of the 157 cases of SSI and sepsis diagnosed, 107 (68%) of the medical records could be located and reviewed. Table 2 shows the demographic profile of all patients who were diagnosed with SSI and sepsis. Only sex and mortality were statistically significantly different (p < 0.05) between patients whose medical records were found, and those whose records were missing (Table 2). Among medical records found, the majority of patients were female (91%) and between the ages of 13-45 years (92%) ( Table 2). Approximately two-thirds of the medical records that were found were of patients admitted in district hospitals (63%). One-third of the patients stayed at the health facility for more than 11 days (34%). Patient diagnoses were evenly distributed between SSIs (45%) and sepsis (postoperative or maternal sepsis), with 14% diagnosed with both. Very few medical records were of patients who died at the facility (3%). Most of the medical records reviewed were of surgical patients (85%), while 15% were of women who delivered vaginally ( Table 2).
Completeness of perioperative documentation: Among surgical patient medical records, 87% included post-operative notes, and among cesarean section patient medical records, 87% included an indication for a cesarean section ( Table 3).
Documentation of SSI and sepsis diagnosis and associated symptoms and vitals: Among the medical records reviewed with SSI cases, a third (34%) of the medical Table 1. Criteria developed to assess the completeness of medical records. Documentation of SSI/sepsis diagnosis Documentation of SSI: The clinician's notes include any combination of the following keywords 'pus draining from the wound,' 'closed wound opened,' 'wound with a foul smell,' and 'wound infection' 1 AND, 1 or more of the following symptoms: heat, redness, localized tenderness, purulent drainage, spontaneous dehiscence, operative findings indicative of infection. Documentation of sepsis: The clinician's notes include a keyword 'sepsis' AND 2 or more of the following criteria: temperature, heart rate, systolic blood pressure, and respiratory rate. We evaluated the presence of recording, not accuracy of measurement.
Patient progress and monitoring Documentation of the following patient progress and monitoring information: daily progress notes documented for each inpatient day*, doctors' orders documented on each inpatient day, partograph present and completed during the perinatal period, discharge details documented (discharge, referral, or death date); post-operative notes documented*, indication for cesarean section documented.
Patient characteristics Documentation of the following patient characteristics: patient history (obstetric admission sheet for obstetric patients), name, age, sex.
*Daily progress notes documented for each inpatient day were written by assistant medical officers/surgeons or another clinical person. Postoperative notes were written by the surgeon directly after the operation.
records included documentation of a diagnosis of SSI and 57% documented signs or symptoms of SSI. 23% of the medical records with sepsis cases documented a diagnosis of sepsis in the medical records, and twothirds (62%) documented two or more vital signs of sepsis (Table 4).
Completeness and utilization of inpatient clinical progress information: Among medical records found, the majority (76%) included patients' history. Less than half (41%) of the daily progress notes were written on all days of admission, and less than onethird (29%) medical records a daily record of doctor's orders. Among obstetric patient records, 79% included a partogram (Table 5).
Overall completeness of medical records: 6% of medical records were complete with documented SSI and sepsis diagnosis, vitals, inpatient clinical progress perioperative documentation indicators, as defined by the research team. We also examined the completeness of medical records among surgical and obstetric patients, and documentation rates were similar (Table 6)

Discussion
Our results show that 32% of the medical records of patients diagnosed with SSI and sepsis were not present at the time of the assessment. Besides missing medical records these findings show that primary data quality is weak in health facilities in northern Tanzania. This is consistent with previous studies examining HMIS in Tanzania at the national level that have also found data quality and reporting to be inefficient and ineffective at the district and national levels [11,13]. Unsatisfactory MTUHA register completion for non-surgical indicators implies that large-scale public health decisions are informed by rough aggregate estimates based on incomplete     data [13]. Our study further highlights this concern by demonstrating that in addition to problems in reporting at the district and national levels, health facilities struggle with collecting, storing, and reporting of patient data at the facility level, which is essential for improving quality of surgical care. Nevertheless, as previous research demonstrates, data quality improvement is a slow process and necessitates a significant change in organizational culture and processes [21]. Despite ongoing efforts, implementation of Electronic Medical Record systems in Tanzania are limited and fragmented, and most facilities depend on paper-based recording of facility-level data [22]. As long as facilities continue to depend on paper records, safe storage and management of patient files are essential. SSIs and sepsis typically develop 5-10 days postoperatively, usually after the patient is discharged from the facility [23][24][25][26]. Studies in high-income countries, as well as LMICs, have found that SSIs are a leading cause of hospital readmissions [27,28]. Thus, timely access to the medical records of returning patients ensures continuity of care at readmission and improves quality patient outcomes. Anecdotally, we found that the reasons for missing medical records were due to inconsistencies in using unique patient identifiers and a lack of storage space for medical records, which is reflective of a lack of infrastructure and human resources to effectively manage medical records.
We found that most facilities do not explicitly document a diagnosis of SSI and sepsis. While documentation rates of vital signs and symptoms are higher, they are still inadequate. Both the explicit documentation of SSI and sepsis, as well as vital signs and symptoms, are essential for improved tracking, diagnosis, and treatment of SSI and sepsis at the facility level. The Ministry of Health of Tanzania mandated that MTUHA be comprehensive in scope. However, thus far, MTUHA has primarily focused on maternal and child health indicators [11]. The absence of mandated surgical indicator reporting on the national level may lead to a lack of accountability at the facility level. One way to improve accountability could be the inclusion of surgical indicators in the MTUHA system to encourage accurate and timely collection, aggregation, and reporting of these indicators to the national level. Improving accurate documentation of surgical indicators is key to effective surveillance, providing facilities with an accurate picture of baseline post-operative complication rates and the existing gaps. Documentation alone does not ensure data accuracy; however, it might improve patient outcomes and is thus the first step towards improving quality of care [29].
Analysis of the completeness and utilization of inpatient clinical progress indicators such as patient history, daily progress notes, doctors' orders, and partogram shows that patient history and partograms were completed and used in most medical records. However, daily progress notes and doctors' orders were not written regularly. This could be due to our conservative measure of completeness for daily progress notes and doctors' orders, which had to be completed daily, for the duration of the patient stay at the facility. We acknowledge stringent data recording processes should not be an end in themselves but instead used to monitor performance, identify gaps in care, provide feedback with the ultimate goal of improve quality of care. However, regular post-operative monitoring is essential to identify early signs and symptoms of SSIs and prevention of surgical complications [30]. Lack of monitoring and documentation of patient progress might lead to patients being discharged too early and without adequate care. Although facilities faced challenges in recording inpatient clinical progress indicators, most medical records included postoperative notes and an indication for cesarean section. Completeness of post-operative notes is crucial for the continuity of care, especially in settings where task-shifting is the primary mode of care delivery, and doctors use post-operative notes to outline treatment plans.
There are a few limitations to this study. Due to the limited resources and difficulty obtaining medical records from the records department, we were unable to retrospectively review the quality of all medical records in our study. The infrastructural capacity varied between the facilities, ranging from wellstaffed and equipped medical records departments to those with facilities lacking space and human resources. For this reason, our study sample consisted of patients who were diagnosed with SSI and sepsis by the medical data collectors as part of the baseline data collection for the study because they were highrisk patients rather than healthy patients whose medical records might be lost due to their short inpatient stay and lack of follow-up. In addition, since the assessment was conducted towards the end of baseline data collection, there is a possibility of a Hawthorne effect since the data collectors were present and observing data collection practices, which might have impacted the surgical teams' behavior.
Finally, the assessment mainly focused on one element of data qualitycompleteness. Other aspects of data quality, including accuracy and reliability, were not evaluated. Tracking data accuracy would require real-time observation of data recording. In our study design, we accounted for a retrospective medical record review as the most cost-effective and efficient means to evaluate data quality.

Conclusion
Quality improvement for safe surgical care is contingent on robust data and consistent record-keeping practices at the facility level. Accurate and reliable medical records provide patient information that can be used to improve continuity of care, communication, and coordination among the surgical providers and ancillary teams as well as hospital managers. Accurate medical records also provide data for decision making for facility-level quality improvement. Furthermore, appropriate reporting of facility-level indicators helps inform evidence-based decision making at the regional and national levels.
Given the far-reaching impact of the quality of medical records for surgical and obstetric patient care, our findings highlight the need to address gaps in primary data quality. Previously implemented strategies to strengthen routine data quality include periodic data quality checks by district MOH staff, rapid feedback to resolve data gaps, and implementation of simple report cards/data dashboards to provide aggregate comparisons of key indicators [31]. Job aids and supervision tools have also been utilized to improve health facility data collection and use [31]. These methods, which have been tested in low-resource settings, could be adapted to the local context and used to improve data completeness in Tanzanian health facilities. Robust record keeping is essential for patient safety and continuity of care, communication between the hospital staff, and the overall quality of care. Surgical quality is multifaceted and notoriously difficult to measure, which further speaks to the importance of accurate surveillance of SSI and sepsis as indicators of surgical quality. The ability to successfully record surgical data is the first step towards measuring and improving quality.

Acknowledgments
We would like to acknowledge and thank all organizations and individuals who contributed to our early consultation on the design of the Safe Surgery 2020 multi-component intervention as well as those organizations and individuals who have contributed to the larger Safe Surgery 2020 initiative including Assist International, Dalberg, Jhpiego, LifeBox, Project ECHO, SPECT, and WFSA. We want to express our sincere appreciation to the Ministry of Health, Community Development, Gender, Elderly and Children (MoHCDGEC), President's Office -Regional Administration and Local Government in Tanzania (PO-RALG), the regional and district medical officers of Mara, Kagera, Shinyanga, Simiyu and Geita regions and the frontline leadership and surgical team members in the study sites for their support of the study.

Author contributions
WL, GM, and SK, designed the study methods and drafted the manuscript. WL, GM, SK, DJ, SM, SA, BN, and NK were involved in the conception and design of the program and evaluation and critically reviewed and approved the final manuscript.

Disclosure statement
All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf and declare: William Lodge II, Salome Kuchukhidze, Gopal Menon, Shehnaz Alidina, Desmond T. Jumbam, and John Varallo had financial support from GE Foundation for the submitted work; John Varallo declares financial support from ELMA Philanthropies. The authors have had no financial relationships with any organizations that might have an interest in the submitted work in the previous three years and no other relationships or activities that could appear to have influenced the submitted work.

Ethics and consent
The study was approved by Harvard Medical School (Boston, MA, USA) and the National Institute for Medical Research (Dar es Salaam, Tanzania). Permissions were sought and obtained from all health facility managers as well as local and regional health managers to conduct the study.

Funding information
This project was funded under grant number 28045607 from the GE Foundation and grant number 17-F0012 from ELMA Philanthropies. The funders will not have any input on data collection, analysis, or interpretation of the data. Further, they will not have any authority over the writing of reports or the decision to submit findings for publication; GE Foundation [28045607].

Paper context
Reliable and complete patient data are essential for surgical quality improvement in low-resource settings. However, the evidence is sparse on the state of paper-based record collection. This study shows significant gaps in medical record availability and completeness for surgical patients diagnosed with surgical site infections and sepsis. Strategies to strengthen routine data quality such as data quality checks, usage of simple report cards, and rapid feedback for process improvement could fill some of these gaps.