13th national convention on statistics (ncs) edsa … 5-3 maternal mortality... · 13th national...

of 28

13th National Convention on Statistics (NCS)

EDSA Shangri-La Hotel, Mandaluyong City

October 3-4, 2016

Maternal Mortality Measurements Using National Surveys and Vital Statistics: Assessing the Quality and Content of Maternal Death Certificates

by

Jeremie E. De Guzman, Manuel M. Dayrit, Percival A. Salting, Anthony Zosa, Jakov Ivan Dumbrique, Camille Dee

For additional information, please contact: Author’s Name Jeremie E. De Guzman Designation Fellow, Ateneo Center for Health Evidence, Action, and Leadership

(AHEALS) Author’s name Manuel M. Dayrit, Percival A. Salting, Anthony Zosa, Jakov Ivan

Dumbrique, Camille Dee Designation Dean, Statistical Specialist II, Professor Affiliation WHO, WHO, DOH, PSA, PSA Address Ateneo Center for Health Evidence, Action, and Leadership, 2nd floor,

Don Eugenio Lopez Sr. Medical Complex, Ortigas Ave., 1604 Pasig City

Tel. No. 531-4151 Email [email protected]

of 28


by

Jeremie E. De Guzman1, Manuel M. Dayrit1, Percival A. Salting2, Anthony Zosa3, Jakov Ivan Dumbrique3, Camille Dee3

ABSTRACT

How well do our vital statistics and national surveys measure Maternal Mortality Ratios (MMR) over time? For the past 25 years, our Maternal Mortality Ratios have remained unchanged. This observation is curious given that so many interventions to improve prenatal care, attendance at delivery by a skilled birth attendant, and post-partum care have been put in place by both national and local governments. In reality, facility-based delivery is now at 80% on average. The question is therefore raised on whether there is in fact no change in MMR or whether the current methods to measure MMR are statistically adequate to detect any changes. To answer the aforementioned question, the authors examined the quality of Vital Statistics 2010-2014 by analyzing the public use files provided by the Philippine Statistics Authority. The authors concluded that data quality was poor based on the low coverage of maternal death reports, the incompleteness of entries, and the lack of internal consistency in the reporting of immediate, antecedent, and underlying causes of death. Thus these could not be relied upon for MMR estimates. On the other hand, national surveys have also been incapable of detecting any changes in MMR over the last 25 years due to the lac of statistical power. In other words, sample sizes were not large enough to detect changes less than +/-40 per 100,000 and to increase them was unaffordable. Furthermore, the inherent limitation of the sisterhood method used in these surveys added to the uncertainty of the estimates of MMR. Moving forward, striving to improve the quality Civil Registry and Vital Statistics is the path which the Philippines should take. This effort should be pursued diligently in the coming years. In the meantime, as far as the MMR baseline for the Sustainable Development Goal is concerned, the country is constrained to use the FHS 2011 figure of 221 per 100,000 as the point of reference for 2030. The country should aim to have reliable maternal mortality estimates from the Vital Statistics within the next few years and certainly before 2030. Keywords: Civil Registry and Vital Statistics, Maternal Mortality Statistics, Maternal Mortality Ratio, Data Quality, Sustainable Development Goals, Demographic Health Surveys, Low–Middle Income Country, Philippines, Sisterhood Method

1 Ateneo de Manila University School of Medicine and Public Health: Ateneo Center for Health Evidence, Action, and

Leadership (A-HEALS). For correspondence please contact Dr Jeremie De Guzman Don Eugenio Lopez Sr. Medical Complex

Ateneo School of Medicine and Public Health, Ateneo Center for Health Evidence, Action, and Leadership, 2nd floor, Ortigas

Ave, Pasig City, Philippines, zip code 1605. Email: [email protected] 2 Philippine Statistics Authority: Demographic and Health Statistics Division 3 Ateneo de Manila University: Department of Mathematics

of 28


by

Jeremie E. De Guzman, Manuel M. Dayrit, Percival A. Salting, Anthony Zosa, Jakov Ivan Dumbrique, Camille Dee

1. Introduction Since 1990, the Philippines’ Maternal Mortality Ratio (MMR) has remained virtually unchanged. The Maternal Mortality Ratio is the primary indicator to monitor progress in achieving Millennium Development Goal (MDG) 5. It is calculated by taking the count of maternal deaths divided by the count of live births multiplied by 100,000. There are two other indicators: the Maternal Mortality Rate that instead has a denominator that uses the number of women in the reproductive age group multiplied by 1,000, and the Proportion of Adult Female Deaths Due to Maternal Causes, which is the number of maternal deaths divided by the reproductive age group. Globally, the gold standard for calculating MMR is by using Civil and Vital Registry Death Certificates to collect those classified as maternal deaths (AbouZahr & Wardlaw, 2001). While this is the case in developed nations, many developing countries lack the information infrastructure, which makes it difficult to monitor progress in reducing maternal mortality. In the Lancet series of 2007 entitled Who counts? it was highlighted that the lack of investments in Civil Registry and Vital Statistics (CRVS) amounted to the failure to obtain adequate information to inform and develop policies to respond to the needs of the people. In the absence of good CRVS databases, other methods were developed to generate estimates on Maternal Mortality. These methods are collectively called the Reproductive Age Mortality Studies (RAMOS). They are generally grouped into five categories: Death Registration, Health Facility Statistics, Decennial Census, Surveys, and Surveillance (see Figure 1 on next page). Outside of the RAMOS, there are other analytic methods to generate Maternal Mortality Ratio estimates.

a. Findings of the 1998, 2006, and 2011 Surveys – no statistical difference between

MMR estimates

There are three sources of Maternal Mortality Data in the country: the Civil Registry and Vital Statistics (CRVS), the National Demographic and Health Surveys (NDHS and its subsets), and the Field Information System (FHSIS). The first CRVS in the Philippines was established in 1898 and institutionalized in 1931 by Act No. 3753. As recently as 2008, new forms have been established to ensure better data granularity, and continuing efforts to migrate the data into digital formats are being pursued by the Philippine Statistics Authority (PSA). The surveys have been done every 7 years since 1998 and are being used to gather data on necessary development indicators and demographic characteristics of the Filipino population. Traditionally, none of these demographic surveys recorded maternal mortality, but specific smaller subset questionnaires of the NDHS were commissioned to generate an estimate for the Maternal Mortality Ratio. The reason these surveys were done was because there was an opportunity to gather Maternal Mortality Data from a larger sample size of households. These were the 2006 Family Planning Survey, the 2011 Family Health Survey. These surveys are now the basis for the official estimates of the Philippine National Maternal Mortality Ratio Indicator. The third source is the Field Health Information System (FHSIS), which generates data from the Public Health Programs of the Department of Health (DOH) at the Rural Health Units established in the 1990s.

of 28

Figure 1. RAMOS Studies from Graham et al., 2008

b. Caveats of the Sisterhood Method

The Direct Sisterhood Method is the method used in the Philippine Surveys. It involves asking living relatives their detailed sibling history to determine if there has been a death in the family that has occurred during pregnancy or what is called a pregnancy related maternal death (PRMD). There are a number of caveats in the use of this method. Prior to even considering using the Sisterhood Method, the World Health Organization (WHO) in 1997 had already cautioned that the use of this survey to generate estimates for MMR was not appropriate for measuring safe motherhood initiatives in the short term, evaluating programme impact, comparing geographic areas (or sub-national estimates), studying trends, or allocating resources. It is important to note that the Sisterhood method relies on the recall of living siblings and does not have any other method of validation. This makes the method inadequate in generating the required information directly relating to the deceased mother for policy and program purposes (Ahmed and Hill, 2010). Another caveat is the sample size required to achieve adequate precision. The 1998 NDHS used a sample of 13,708 households, the FPS a sample of 45,000 households, and the FHS a sample of 49,000 households. Even with the larger sample sizes used in the FPS and FHS surveys, the generated confidence intervals for the MMR estimates were still very wide at +/- 40%, as shown in Figure 2 and Table 1. In Figure 2 each of the point estimates are different: 172/100k, 162/100k, and 221/100k. Even if there have been differences in point estimates in the MMR, if the changes are not larger than 40%, then any changes will not be considered statistically different from each other. This is reflected in Figure 2 as the bands of all 3 data points overlap with each other. In addition, the Sisterhood Method should not be used in countries that have low total fertility rates (i.e., less than 3) and high internal migration. It is not recommended to repeat the survey less than 10 years after one round because of its low precision and retrospective nature (World Health Organization [WHO] Department of Health and Research, 1997). Low Fertility rates and high migration patterns affect the Sisterhood Method in the same direction. This means that they both decrease the probability of detecting a maternal death thereby making

of 28

a rare event harder to measure. The net effect is to require larger sample sizes to narrow the confidence intervals of the estimates. Lastly, it is important to note the recommended 10-year period gap between surveys. The MMR represents a magnitude/range of obstetric risk estimate spread over a number of years (typically 7 years according to Hill et al., 2006). If the surveys are repeated in a period of less than the recommended years, changes in maternal health may not yet reflect in the MMR estimates produced from the survey.

Table 1

NDHS, FPS, and FHS MMR Estimates with the Range and Confidence Intervals

NDHS 1998, FPS 2006 and FHS 2011 Maternal Mortality Ratio Estimates

Year Average Lower-Upper Range 95% Confidence Interval

1998 172 132-212 +/- 40

2006 162 122-202 +/- 40

2011 221 181-261 +/- 40

Figure 2 Boxplot of Each National Survey with Adult Maternal Mortality Ratios, Taken

from the 2011 Family Health Survey Report

It is clear that using the survey as a method to monitor progress towards reducing Maternal Mortality is inadvisable given its limitations. As a proxy, the United Nations (UN) and WHO have recommended using the operational indicators of Skilled Birth Attendance at Birth (SBA), Facility-Based Delivery (FBD), Antenatal Care (ANC), and Post-Partum Care (PPC). Collectively, these are known as the UN process indicators, and they are much easier to collect and interpret for program management given the difficulties in estimating and interpreting the MMR. However the relationship of High coverage of the UN process indicator leading to better maternal health outcomes is not straightforward (Souza et al. 2014). Shifting to the Sustainable Development Goals, AbouZahr and colleagues in 2015 have stated in the Lancet that the time is ripe to revisit strengthening the CRVS as it is the most reliable and cost-efficient data source that a country can have with regards to mortality.

of 28

c. The Death Certificate and Relevant Data entered for Analyzing Maternal Mortality

Figure 3. Sample Death Certificate from the Civil and Vital Registry vr 2008

The Data generated by the death certificates are very important. If properly filled it can answer who died, where they died, how they died, and whether the death was from natural or other causes. Maternal Mortalities are a subset of all mortalities reported in the death certificates. The important data fields that can shed light on the circumstances surrounding a maternal death include: (1) Demographic Data (2) Causes of Death (3) Attendant at Death (4) Maternal condition or maternal timing of death. In a death certificate there are 3 data field entries for cause of death, namely: Immediate, Antecedent, and Underlying. Any mortality is classified based on the underlying cause of death (Handbook for Medical Certification of Death, DOH 2014). For example if the underlying cause of death is tuberculosis, said death is classified as a TB death. Thus a maternal mortality is identified if the underlying cause of death is obstetric in nature. II. Methods

In partnership with the Philippine Statistics Authority, the Ateneo team was given access to specific data fields from the CRVS public use files on maternal death certificates from 2010 to 2014. The team was also given access to the raw data and report of the Family Health Survey 2011, and the final report of the National Demographic and Health Survey 2013. Using PSPP, frequency tables were generated from the CRVS to count the reported deaths, Age distribution of those who died, the regions where they died, the immediate causes of death, the antecedent causes of death, the underlying causes of death, the attendant at death, and the maternal condition or timing at death.

of 28

Microsoft excel was used to filter and cross tabulate records that had complete entries in all fields from the data sets as well as generating summations for different variable values. It was also used to generate 10 random entries from the 2014 data set with complete entries on all fields.

III. Results

a. The Estimated Coverage of Death Certificates in the Philippines is Low,

affecting overall data quality

The estimated coverage of death certificates being reported were at 65% in 2009. This percentage was estimated using the Preston and Coale method developed in 1981. To calculate this, the data requirements are to collect the census over a specific reference period (in this case the total population in the 2000 Census of the Philippines) and then to estimate the number of mortalities using an equation in reference to the current year of interest (2009). The total number of death certificates submitted is then divided by the estimated total mortalities of the period (Cajita, Tan, et al. 2009).

b. MMR estimates from CRVS 2010 to 2014 demonstrate some fluctuations up and down but the general trend is flat relative to the last 25 years

Table 2 Total Number of Reported Deaths, Live Births, and MMR Using the CRVS

CRVS MMR

Year

Reported Deaths Reported Live Births MMR

2010

1,719 1,782,981 96.41

2011

1,469 1,746,684 84.10

2012

1,447 1,790,367 80.82

2013

1,522 1,761,602 86.40

2014

1,765 1,748,857 100.92

Figure 4. MMR Trend from 2010 to 2014 Using the CRVS Data

of 28

From 2010 to 2014, the estimated Maternal Mortality Ratio using the CRVS is 96 to 100 deaths per 100,000. As shown in the figure, there are yearly fluctuations in the MMR over the past 5 years however general trends in either direction are not sustained. Since the 1980s the Philippine CRVS reports MMRs ranging from 80 to 110/100,000 and has remained relatively stagnant. These figures and trends are reflected in the Trends in Maternal Mortality Report by the UN, WHO, and UNICEF in 2013. This report used the same CRVS data to compare the progress for achieving MDG 5 across different countries to see progress over the 25-year period. For the Philippines the report estimated that the percent change over this period was 25% with an average annual fluctuation of 1.1%. The report then classified the Philippines as one of 26 countries that had no progress in reducing the MMR by 70%.

c. Mean Age Distribution of Deaths is 30 years old with a standard deviation of +/- 7

years

Table 3 Year and Age of Those who Died

Year Reported Deaths Variable Mean Std Dev Range Minimum Maximum

2010 1719 Age 30.17 7.55 47 13 60

2011 1469 Age 30.49 7.97 77 14 91

2012 1447 Age 30.66 7.78 34 15 49

2013 1522 Age 30.12 7.61 34 15 49

2014 1765 Age 30.2 7.56 34 15 49

From the CRVS, the mean age of mothers who died is 30 years old with a standard deviation of 7-8 years. Thus the age distributions of Maternal Deaths are clustered around the 23-38 years old range. There may be some misclassified deaths as the data from 2010 and 11 reported the deaths of a 60 year old and a 91 year old and both of these ages are beyond the reproductive age group population.

d. Maternal Deaths are higher in Regions with high poverty incidence

of 28

Figure 4. Intensity of Death Reports by Region

From 2010 to 2014 majority of reported deaths come from the following regions: NCR, CALABARZON and MIMAROPA, Central Visayas, Western Visayas, Bicol, and Central Luzon. Citing the year 2012 as an example the highest counts for maternal deaths are in regions with relatively higher poverty incidence rates, except for NCR (Table 4).

of 28

Table 4.

2012 Reported Deaths, Urbanization Levels, and Poverty Incidence

Region Name Region Code

Reported Deaths

Percentage Urban Barangays Percentage

Poverty Incidence Estimates

NCR 13 188 13.00 100% 2.60%

CALABARZON and MIMAROPA

17 180 12.00 59.7/22.3% 31.90%

Bicol 5 130 9.00 15.30% 32.30%

Central Visayas 7 123 9.00 43.70% 25.70%

Central Luzon 3 121 8.40 51.60% 10.10%

Eastern Visayas 8 110 7.60 8.70% 37.40%

Wester Visayas 6 99 6.80 34.70% 22.80%

Northern Mindanao 10 82 5.70 41.30% 32.80%

Davao Region 11 71 4.90 59.30% 25%

Ilocos Region 1 63 4.40 12.70% 14%

Zamboanga Peninsula 9 61 4.20 33.90% 33.70%

Soccsksargen 12 59 4.10 46.50% 37.10%

NIR 18 52 3.60 * *

Caraga 16 46 3.20 27.50% 31.80%

Cagayan valley 2 37 2.60 11.60% 17%

CAR 14 19 1.30 26.30% 17.50%

ARMM 15 6 0.40 13.70% 48.70%

e. The top underlying causes of death are hypertensive diseases and

hemorrhage

of 28

Table 5.Underlying Causes of Death Table 2010 to 2014 from the ICD 10 Codes

2010 2011 2012 2013 2014

O94-O99: Other obstetric conditions, Not elsewhere classified

6% 6% 5% 4% 4%

O85-O92: Complications Predominantly related to Puerperium

9% 7% 10% 11% 11%

O80-O84: Encounter of Delivery 0% 0% 1% 1% 1%

O72: Post Partum Hemorrhage 17% 17% 15% 14% 16%

O60-O75: Complications of Labour and Delivery (excluding O72: Post Partum Hemorrhage)

17% 19% 17% 16% 15%

O30-O48: Maternal Care Related to the Fetus and Amniotic Cavity, and Possible Delivery Problems

7% 11% 6% 7% 7%

O20-O29: Other Maternal Conditions predominantly related to Pregnancy

0% 1% 1% 1% 1%

O10-O13 and O16: Other Edema, proteinuria, and Hypertensive disorders of pregnancy, childbirth and the puerperium

1% 0% 1% 1% 2%

O15: Eclampsia 22% 20% 20% 20% 20%

O14: Pre-eclampsia 12% 10% 14% 16% 14%

O00-O08: Pregnancy with abortive outcome 9% 8% 10% 10% 11%

Total 100% 100% 100% 100% 100%

Hypertensive diseases of pregnancy (Eclampsia, Pre-eclampsia) and post partum hemorrhage have consistently been cited as the top causes of mortality. These two comprise 50% of all cited causes of maternal deaths.

f. Most of the Death Reports come from Hospital Authorities

of 28

Figure 5. Attendant at Death

For the past five years, the percentage of maternal deaths that were attended by “hospital authorities” average at 66%, by “public health officers” at 3.7%, and by “private physicians” at 1.2%. Those that were not attended average at 28%. This indicates that many of the deaths are reported come from hospitals.

g. Over half of Reported Maternal Deaths have “None of these” with regards to

their Timing.

Figure 6. Timing of Death

The maternal condition or the timing of maternal deaths classify when a death occurred in the pregnancy period. Prior to 2013 and 2014, none of the maternal death certificates had any

of 28

entries in this field. More than 50% of the entries in this field were listed as “None of these” for the two years where the data was inputted.

h. The Circumstances surrounding a maternal death are often unclear as reported in

the death certificate

As a general observation from 2010 to 2014 completeness entry improved slightly. In 2013 8.3% had entries for all three causes of death and maternal condition/timing. In 2014 13.4% had entries for all three fields and maternal condition/timing. From the completed cases 10 were randomly chosen in an attempt to understand the circumstances behind the reported deaths (Table 5). Many of the cases appeared to confuse mode or mechanism of death with the cause of death. In addition many of the cases also had ill-defined cause of death codes and some of the cases confused the sequence of events leading to the death. In all, this makes interpreting the circumstances behind each death even more difficult.

of 28

Table 6

Randomly Selected Cases of Causes of Death and Timing of Deaths Comparison 2014

Immediate Antecedent Underlying Timing of Death Attendant at Death Remarks

I469: Cardiac arrest, unspecified

R571: Hypovolemic shock

O066: Unspecified abortion

Pregnant not in labor

None Plausible but ill-defined underlying cause of death

R571: Shock, not elsewhere classified

O721: Other immediate post-partum hemorrhage

O622: Secondary uterine inertia

Less than 42 days after delivery

Hospital authority

Plausible

O141: Severe preeclampsia

O459: Premature separation of placenta, unspecified

O622: Other uterine inertia

Pregnant in labor Hospital authority

Plausible but order of events is incorrect


I10X: Hypertension

O152: Eclampsia in puerperium


Plausible but maternal timing and underlying code are contradicting

I279: Pulmonary heart disease, Unspecified

G934: Encephalopathy, unspecified

O622: Other uterine inertia

42 days to 1 year after delivery

Hospital authority

Implausible given timing and causes of death


A419: Sepsis, unspecified

O723: Post-partum coagulation defects


Hospital authority

Implausible given the chain of events

I461: Sudden cardiac death

I219: Acute myocardial infarction, unspecified

O994: Diseases of the circulatory system complicating pregnancy


Hospital authority

Plausible chain of events

G931: Anoxic brain damage, not otherwise specified

I509: Heart failure, unspecified

O903: Cardiomyopathy in puerperium


Hospital authority


I269: Pulmonary embolism without mention of acute cor pulmonale

O149: Preeclampsia, unspecified

O459: Premature separation of placenta, unspecified



D649: Anaemia, unspecified

K922: Gastrointestinal hemorrhage, unspecified

O159: Eclampsia, unspecified to time period

Pregnant in labor None Difficult to determine

of 28

IV. Discussion

a. Hypertensive Diseases of Pregnancy and Post Partum hemorrhage are still the leading causes of Maternal Deaths

It is alarming that the top causes of maternal deaths are due to Hypertensive diseases of pregnancy and Post Partum Hemorrhage. Even though these are condition with which the maternal program has been designed to prevent and treat. The NDHS in 2013 provides clues on why this may have happened. Using the survey it is a consistent trend that poorest and least educated women are given less of the services (USAID and UPEcon 2016). This may also be compounded by operational issues that hamper the delivery of necessary supplies as well as issues in the quality of the services being given. Thus it is important to relate whether the poor quality of services and operational issues occur in the areas where there are a high number of reported deaths.

b. Maternal Death Counts are higher in regions with higher poverty incidence

Except for NCR many of the regions that report a higher number of maternal deaths also have higher poverty incidences. This observation supports the literature that poverty affects overall health and contributes to mortality. NCR is an exemption to this but it is possible that given its population density and the availability of end-referral facilities in the capital region that many of the deaths are reported here. For example, if a mother from the poorer regions needed more advanced care, she would be referred to hospitals in the capitol and if she died she would be reported as a death from NCR though she did not necessarily live there. This phenomenon is unproven but would require further investigation given that NCR has the lowest poverty incidence but one of the highest numbers of maternal deaths reported. Analyzing the relationship between poverty incidence and the number of reported deaths in subnational levels should be pursued in order to evaluate the mechanisms and contexts where these deaths were reported to have occurred.

c. Overall Data Quality of CRVS 2010 to 2014 is poor

At present the overall data quality of the CRVS is poor. The two primary contributors to why would be the low coverage of reported deaths and the poor quality of data inputted. The estimated coverage of 65% is considered to be low. Other countries with well functioning CRVS systems have coverage of 80% or more and are routinely used to inform policy and operations. Evidence of poor quality can be seen from the misclassified deaths in 2010 and 2011 based on age, the incomplete entries in the causes of death, and timing of death. Overall only 13.4% of death certificates are completely filled. This calls into question the utility and reliability of the data for policy and operation use. As the surveys, by themselves, do not provide enough information to inform policy and operations a systematic and sustained effort in improving the CRVS should be implemented and supported as it is a crucial tool in informing and monitoring efforts in reducing maternal mortality.

d. Improving CRVS is the most cost effective way of measuring maternal mortality.

The CRVS is the single most reliable and cost effective means of measuring maternal mortality (AbouZahr et al. 2015). The two main issues in order to improve the CRVS in the Philippines are (1) the low coverage of death certificate reports and (2) the accuracy of the causes of maternal death and the timing of maternal death.

of 28

Mikkelsen and colleagues in 2015 proposed that a new measure, the Vital Statistics Performance Index (VSPI). It uses mortality data as a proxy metric for the quality and utility of all the CRVS data produced. It uses a single scale from 0 to 1 for each calendar year produced. Values closer to 1 approximate the real epidemiologic profile of the country and indicate fitness for policy use, whereas values closer to zero indicate those which are unrepresentative and that offer little use for policy making. The six components of this metric are (1) completeness/coverage of death reporting, (2) quality of death reporting, (3) level of cause-specific detail (4) internal consistency, (5) quality of age and sex reporting, and (6) data availability or timeliness. Improving completeness is the most complex and time consuming of all six, but it is also the most cost-effective. For example, strategic initiatives to train doctors in correct medical certification, adoption, and use of the standard international cause of death certificate and use of more detailed cause of death lists can improve the quality of CRVS in a decade, as was observed in other countries. According to the study the Recent Philippine VSPI scores calculated for 2005 to 2010 and 2011 to 2012. These were at 45.6% and 20% respectively. This means that the CRVS data must be improved to increase its utility in policy and operations. Improving the CRVS data hinges on addressing the two main issues identified – i.e., increasing coverage of death certificate reports in the Philippines and better accuracy of the data entered in the certificates. It has been postulated that requiring funeral homes and mortuaries to report deaths to authorities can increase coverage substantially. The Philippines already has existing policies to allow this and revisiting its implementation with local government and the Department of Health can be a potential way to increase coverage rapidly (AbouZahr et al., 2015). Implementing the use of modern communications technology such as SMS or the internet to inform local authorities of mortality is also potentially cost-efficient means of increasing the coverage of reported deaths. Another potential source of information would be Philhealth Claims Data, especially if these can be matched to the reported mortalities. Lastly current coverage also covers mostly the public sector, and involving more private sector facilities and clinics in reporting deaths should be pursued. Efforts in improving accuracy and completeness of entry for death certificates are underway. A handbook was co developed by the PSA and the DOH as a reference for physicians in the field. Changes in medical education where also introduced. Lessons from other countries can also be adopted. For example, in Thailand including coverage of registration as a component of good governance metrics with corresponding incentives was able to increase coverage. (Thai Health Information Standards Development Center 2013)

e. As a baseline for the SDG, use the FHS 2011 figure of 221/100,000 MMR

As a baseline for the SDGs the Philippines should use 221/100,000 MMR figure from the FHS 2011. As a broad starting point it appears that this number is more reliable compared to the data generated from the CRVS at present. Moving forward the CRVS should be strengthened to be able to generate the MMR reliably. Even with its limitations the CRVS is better suited to monitor annual and subnational changes in the MMR compared to the national surveys.

of 28

In the interim, it is prudent to use the different sources of data to create a bigger picture of the maternal mortality situation. For instance the NDHS 2013 was able to demonstrate inequity in both health access and the impact of the social determinants of health.

f. Observations from the 2013 NDHS

Looking at the quintile breakdowns, the national averages hide inequities in access to maternal health services. Figure 7 compares the 2013 averages across the quintiles of the Philippine population. It can be observed that there are large discrepancies between the lowest quintile and the others, especially on SBA, FBD, and PPC. A more thorough investigation would be warranted from the existing NDHS on the subnational distribution and magnitude of these inequities as they may be reason for the relative stagnation of Maternal Mortality Ratios over the past 25 years.

It was demonstrated that aside from the poor having the lowest coverage and completeness of services, they are also the least educated. In other words the poorest and least educated get the least amount and quality maternal health services. (USAID, UPEcon 2016)

Figure 7. Quintile Breakdown of the Maternal Health indicators NDHS 2013

Relating this to the findings of the CRVS, it is possible that even in urbanized locations, the relatively high proportion of poverty and the large impact of a single maternal mortality (since it is a rare event) exponentially increases the MMR. In light of this, the inequity of access for the poorest quintile and the relatively high number of maternal death reports in regions with above average poverty incidences point to a potentially large impact. Further research should be done to investigate this discrepancy. Other factors such as the social determinants of health are important considerations in improving the effectiveness of the Maternal Health Program moving forward. Among all the 144 Low-Middle Income Countries that have embarked on reducing maternal mortality 10 countries which have achieved success showed that there is no one template in reducing maternal mortality but that it is important to integrate health sector interventions with other women initiatives that address the

0

0.2

0.4

0.6

0.8

1

1.2

Lowest Second Middle Fourth Fifth

2013

2013 Maternal Health Process Indicators across Quintiles

ANC

SBA

FBD

PPC

of 28

underlying social determinants such as livelihood, education, and governance (Kuruvilla et al. 2014). Shifting to the Sustainable Development Goals (SDG) would require constant commitment not only in terms of the interventions to be implemented but also on using and improving data for monitoring and evaluation purposes. V. Conclusion

The Philippines has mainly relied on the Sisterhood Method to generate its MMR point estimates. Over the last 25 years, each of these estimates has not shown statistically significant differences between one another. Given the large confidence intervals associated with the national surveys, it cannot be ascertained whether there have indeed been changes in Maternal Mortality Rates. The CRVS is seen as the gold standard for monitoring the vital events in a country, and it is also used to gauge development. The Philippine CRVS is currently at 51% usability according to the VSPI, and moving forward, it should be used to monitor maternal mortality. In addition, continuous improvements and strengthening is imperative to increase the utility of the CRVS, in particular, in terms of the coverage of death certificate reporting and the accuracy of inputted data. Despite the shortcomings, the CRVS still reveals important patterns in maternal mortality at a glance. Combining it with the current NDHS 2013 report, it can be postulated that the high death rates in the lowest quintiles may have the effect of keeping the overall MMR high. Shifting to the SDGs, the current 2011 MMR estimate from the FHS can be used as a starting point. However, the surveys should not be used to monitor progress annually and are inappropriate for the annual monitoring of the MMR. The CRVS, even in its current state, is better suited to guide program direction and monitor progress annually and over different geographic locations and jurisdictions.

Table 6 Strengths and Weaknesses of Selected Typology of Data Opportunities and Options for Measuring Maternal

Mortality (Adapted from Graham et al., 2008)

Data Opportunity Strengths Limitations

Death Registration - Philippine Civil and Vital

Statistics System

- Also collects birth data - Representative of entire

population - Continuously available - Can provide annual and cause-

specific estimates of maternal mortality, trends, and sub-national distribution

- Under-reporting - Untimely/late reporting of

deaths and births - Quality of cause-of-death

certification and coding needs regular assessment

Health Facility Statistics - Field Health Information System

of the Philippine Department of Health

- Ongoing routine data collection at facilities

- Provides annual and cause-specific and trend estimates of maternal mortality

- Captures only facility deaths (specifically rural/city health units) and misses community/non-obstetric ward deaths

- Records may be poorly maintained

Surveys - National Demographic and

Health Survey (NDHS) and its subsets, the Family Planning

- Data representative of the population

- Generates a more recent estimate vs. the indirect method

- Larger sample size than indirect method with more complex collection, processing, and analysis

of 28

Survey (FPS) and the Family Health Survey (FHS), using the Direct Sisterhood Method

(which reflects the situation 20 years prior to collection)

- Permits calculation of confidence intervals and can be added to multi-purpose surveys

- Required sample size to achieve precision to detect changes in MMR is not cost-effective

- Less appropriate with substantial migration

- Should have a 10 year gap between surveys

- No established demographic techniques to permit evaluation and adjustment, if necessary

of 28

References AbouZahr, C., & Wardlaw, T. (2001). Maternal mortality at the end of a decade: Signs of progress?.

Bulletin of the World Health Organization, 79(6), 561–568. http://doi.org/10.1590/S0042-96862001000600013

AbouZahr, C., Cleland, J., Coullare, F., Macfarlane, S. B., Notzon, F. C., Setel, P., & Szreter, S. (2007). The way forward. The Lancet, 370(9601), 1791–1799. http://doi.org/10.1016/S0140-6736(07)61310-5

AbouZahr, C., De Savigny, D., Mikkelsen, L., Setel, P. W., Lozano, R., & Lopez, A. D. (2015). Towards universal civil registration and vital statistics systems: the time is now. The Lancet, 386(10001), 1407–1418. http://doi.org/10.1016/S0140-6736(15)60170-2

AbouZahr, C., De Savigny, D., Mikkelsen, L., Setel, P. W., Lozano, R., Nichols, E., . . . Lopez, A. D. (2015). Civil registration and vital statistics: progress in the data revolution for counting and accountability. The Lancet, 386(10001), 1373–1385. http://doi.org/10.1016/S0140-6736(15)60173-8

Ahmed, S., & Hill, K. (2016). Maternal mortality estimation at the subnational level: a model-based method with an application to Bangladesh. Bulletin of the World Health Organization, 89(), 12–21. http://doi.org/10.2471/BLT.10.076851

Ahmed, S., Li, Q., Scrafford, C., & Pullum, T. W. (2014). An Assessment of DHS Maternal Mortality Data and Estimates. DHS Methodological Reports, 13(September), 1–140. http://dhsprogram.com/pubs/pdf/MR13/MR13.pdf

Alam, N., & Townend, J. (2014). The Neighbourhood Method for Measuring Differences in Maternal Mortality, Infant Mortality and Other Rare Demographic Events. PLoS ONE, 9(1). http://doi.org/10.1371/journal.pone.0083590

Bhutta, Z. A., & Black, R. E. (2013). Global Maternal, Newborn, and Child Health — So Near and Yet So Far. The New England Journal of Medicine, 369(23), 2226–2235. http://doi.org/10.1056/NEJMra1111853

Bhutta, Z. A., Hafeez, A., Rizvi, A., Ali, N., Khan, A., Ahmad, F., . . . Jafarey, S. N. (2013). Reproductive, maternal, newborn, and child health in Pakistan: challenges and opportunities. The Lancet, 381(9884), 2207–18. http://doi.org/10.1016/S0140-6736(12)61999-0

Campbell, O. M. R. (2001). What Are Maternal Health Policies in Developing Countries and Who Drives Them? A Review of the Last Half-century. Retrieved from http://www.jsieurope.org/safem/collect/safem/pdf/s2947e/s2947e.pdf

Curtis, S. L., Mswia, R. G., & Weaver, E. H. (2015). Measuring Maternal Mortality: Three Case Studies Using Verbal Autopsy with Different Platforms. PLoS ONE, 10(8), 1–16. http://doi.org/10.1371/journal.pone.0135062

Department of Health Knowledge Management and Information Technology Service. (2015). Medical Certification of Death: Handbook for Filipino Physicians. Retrieved from http://www.doh.gov.ph/sites/default/files/basic-page/MEDICAL_CERTIFICATION_OF_DEATH_0.pdf

Fauveau, V., & Donnay, F. (2006). Can the process indicators for emergency obstetric care assess the progress of maternal mortality reduction programs? An examination of UNFPA Projects 2000–2004. International Journal of Gynecology and Obstetrics, 93(3), 308–316. http://doi.org/10.1016/j.ijgo.2006.01.031

Firoz, T., Chou, D., Von Dadelszen, P., Agrawal, P., Vanderkruik, R., Tunçalp, O., . . . Say, L. (2013). Measuring maternal health: focus on maternal morbidity. Bulletin of the World Health Organization, 91(10), 794–796. http://doi.org/10.2471/BLT.13.117564

Graham, W. J., Ahmed, S., Stanton, C., Abou-Zahr, C. L., & Campbell, O. M. R. (2008). Measuring maternal mortality: An overview of opportunities and options for developing countries. BMC Medicine, 6(12). http://doi.org/10.1186/1741-7015-6-12

of 28

Hanley, J. A., Hagen, C. A., & Shiferaw, T. (1996). Confidence Intervals and Sample-size Calculations for the Sisterhood Method of Estimating Maternal Mortality. Studies in Family Planning, 27(4), 220–227. http://doi.org/10.2307/2137955

Hill, K., El Arifeen, S., Koenig, M., Al-Sabir, A., Jamil, K., & Raggers, H. (2006). How should we measure maternal mortality in the developing world? A comparison of household deaths and sibling history approaches. Bulletin of the World Health Organization, 84(3), 173–180. http://doi.org/10.2471/BLT.05.027714

Hill, K., Lopez, A. D., Shibuya, K., & Jha, P. (2007). Interim measures for meeting needs for health sector data: births, deaths, and causes of death. The Lancet, 370(9600), 1726–1735. http://doi.org/10.1016/S0140-6736(07)61309-9

Hill, K., Stanton, C., & Gupta, N. (2001). Measuring Maternal Mortality from a census: guidelines for potential users. MEASURE Evaluation Manual Series, 4(July), 1–53. http://www.who.int/maternal_child_adolescent/documents/measuring_maternal_mortality.pdf

Horton, R. (2007). Counting for health. The Lancet, 370(9598), 1526. http://doi.org/10.1016/S0140-6736(07)61418-4

Hufana, L., Cajita, J., Morante, L., Lopez, J., Tan, C. L., Mikkelsen, L., & Aung, E. (2009). Assessing the production, quality and use of national vital statistics: a case study of the Philippines. Health Information Systems Knowledge Hub, 2(November), 1–26. http://www.nsor12.ph/pdf_files/CIVIL%20REGISTRATION%20LAWS/Assessing%20Vital%20Stat%20in%20the%20Philippines%20by%20NSO%20CRD.pdf

Idris, H., Tyoden, C., Ejembi, C., & Taylor, K. (2010). Estimation of Maternal Mortality using the Indirect Sisterhood Method in Three Communities in Kaduna State, Northern Nigeria. African Journal of Reproductive Health, 14(3), 77–82. http://www.bioline.org.br/pdf?rh10052

Kassebaum, N. J., Bertozzi-Villa, A., Coggeshall, M. S., Shackelford, K. A., Steiner, C., Heuton, K. R., . . . Lozano, R. (2014). Global, regional, and national levels and causes of maternal mortality during 1990–2013: a systematic analysis for the Global Burden of Disease Study 2013. The Lancet, 384(9947), 980–1004. http://doi.org/10.1016/S0140-6736(14)60696-6

Kijsanayotin, B., Ingun, P., & Sumputtanon, K. REVIEW OF NATIONAL CIVIL REGISTRATION AND VITAL STATISTICS SYSTEMS : A case study of Thailand (2013). http://www.this.or.th/files/77.pdf

Kirby, T. (2015). Alan Lopez: Making everyone count. The Lancet, 386(10001), 1331. http://doi.org/10.1016/S0140-6736(15)60506-2

Kuruvilla, S., Schweitzer, J., Bishai, D., Chowdhury, S., Caramani, D., Frost, L., . . . Bustreo, F. (2014). Success factors for reducing maternal and child mortality. Bulletin of the World Health Organization, 92(7), 533–544. http://doi.org/10.2471/BLT.14.138131

Lo, S., & Horton, R. (2015). Everyone counts—so count everyone. The Lancet, 386(10001), 1313–1314. http://doi.org/10.1016/S0140-6736(15)60305-1

Lopez, A. D., AbouZahr, C., Shibuya, K., & Gollogly, L. (2007). Keeping count: births, deaths, and causes of death. The Lancet, 370(9601), 1744–1746. http://doi.org/10.1016/S0140-6736(07)61419-6

Mahapatra, P., Shibuya, K., Lopez, A. D., Coullare, F., Notzon, F. C., Rao, C., & Szreter, S. (2007). Civil registration systems and vital statistics: successes and missed opportunities. The Lancet, 370(9599), 1653–1663. http://doi.org/10.1016/S0140-6736(07)61308-7

Mammaro, A., Carrara, S., Cavaliere, A., Ermito, S., Dinatale, A., Pappalardo, E. M., . . . Pedata, R. (2009). Hypertensive Disorders of Pregnancy. Journal of Prenatal Medicine, 3(1), 1–5. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3279097/

Mikkelsen, L., Phillips, D. E., AbouZahr, C., Setel, P. W., De Savigny, D., Lozano, R., & Lopez, A. D. (2015). A global assessment of civil registration and vital statistics systems: monitoring data quality and progress. The Lancet, 386(10001), 1395–1406. http://doi.org/10.1016/S0140-6736(15)60171-4

Mikkelsen, L., Rampatige, R., Hernandez, B., & Lopez, A. D. (2014). Strengthening vital statistics systems: What are the practical interventions necessary to reduce ignorance and uncertainty about causes of death and disease burden in the Asia Pacific region? Asia Pacific Observatory on

of 28

Health Systems and Policies: Policy Brief, 3(2), 57. http://www.wpro.who.int/asia_pacific_observatory/resources/policy_briefs/policy_brief_strengthening_vital_statistics_systems.pdf

Okonjo-Iweala, N., & Osafo-kwaako, P. (2007). Improving health statistics in Africa. The Lancet, 370(9598), 1528–1529. http://dx.doi.org/10.1016/S0140-6736(07)61644-4

Paxton, A., Bailey, P., & Lobis, S. (2006). The United Nations Process Indicators for emergency obstetric care: Reflections based on a decade of experience. International Journal of Gynecology and Obstetrics, 95(2), 192–208. http://doi.org/10.1016/j.ijgo.2006.08.009

The Philippines Civil Registration and Vital Statistics Strategic Plan. (2014). Philippines National Statistical Coordination Board. (2013). 2012 Full Year Official Poverty Statistics, 1–

75. Phillips, D. E., AbouZahr, C., Lopez, A. D., Mikkelsen, L., De Savigny, D., Lozano, R., . . . Setel, P. W.

(2015). Are well functioning civil registration and vital statistics systems associated with better health outcomes? The Lancet, 386(10001), 1386–1394. http://doi.org/10.1016/S0140-6736(15)60172-6

Say, L., & Raine R. (2007). A systematic review of inequalities in the use of maternal health care in developing countries. Bulletin of the World Health Organisation, 85(10), 812–819. http://www.ncbi.nlm.nih.gov/pubmed/18038064

Setel, P. W., Macfarlane, S. B., Szreter, S., Mikkelsen, L., Jha, P., Stout, S., & AbouZahr, C. (2007). A scandal of invisibility: making everyone count by counting everyone. The Lancet, 370(9598), 1569–1577. http://doi.org/10.1016/S0140-6736(07)61307-5

Shibuya, K., & Gilmour, S. (2015). Civil registration as a means to promote human security. The Lancet, 386(10001), e14–e15. http://doi.org/10.1016/S0140-6736(15)60765-6

Souza, J. P., Gülmezoglu, A. M., Vogel, J., Carroli, G., Lumbiganon, P., Qureshi, Z., … Say, L. (2013). Moving beyond essential interventions for reduction of maternal mortality (the WHO Multicountry Survey on Maternal and Newborn Health): A cross-sectional study. The Lancet, 381(9879), 1747–1755. http://doi.org/10.1016/S0140-6736(13)60686-8

Sta. Maria, M. O. (1993). The Field Health Services Information System: Its Role in Decentralizing Health Services in the Philippines, 4. http://pdf.usaid.gov/pdf_docs/PNABR934.pdf

Stanton C. Abderrahim N, and Hill K. Demographic and Health Surveys Analytical Report No. 4. September 1997. Maryland, USA

USAID and UPecon Foundation. (2016). Reaching the Poor with a Continuum of Care: A 25-Year Assessment of Health Sector Performance [Draft].

WHO, UNICEF, UNFPA, World Bank Group, and the United Nations Population Division. (2015). Trends in maternal mortality: 1990 to 2015. Retrieved from http://apps.who.int/iris/bitstream/10665/194254/1/9789241565141_eng.pdf?ua=1

World Health Organization. (2014). Civil Registration and Vital Statistics Investment Planning: Report of a technical consultation, Addis Ababa, 28–29 April 2014, 1–26. Retrieved from http://www.who.int/healthinfo/civil_registration/crvs_meetingreport_april2014.pdf?ua=1

World Health Organization. (2016). Making every baby count: audit and review of stillbirths and neonatal deaths. 144. Retrieved from http://apps.who.int/iris/bitstream/10665/249523/1/9789241511223-eng.pdf?ua=1

World Health Organization Division of Reproductive Health and Research. (1997). The sisterhood method for estimating maternal mortality: Gudiance notes for potential users, 1–22. Retrieved from http://apps.who.int/iris/bitstream/10665/64007/1/WHO_RHT_97.28.pdf

Van Lerberghe, W., Matthews, Z., Achadi, E., Ancona, C., Campbell, J., Channon, A., . . . Turkmani, S. (2014). Country experience with strengthening of health systems and deployment of midwives in countries with high maternal mortality. The Lancet, 384(9949), 1215–1225. http://doi.org/10.1016/S0140-6736(14)60919-3

of 28

Villar, J., & Pérez-Méndez, L. (2007). Evaluating an educational intervention to improve the accuracy of death certification among trainees from various specialties. BMC Health Services Research, 7(183). http://doi.org/10.1186/1472-6963-7-183

13th national convention on statistics (ncs) edsa … 5-3 maternal mortality... · 13th national...

Documents