Bas-relief at Angkor Wat, Cambodia, c. 1150, depicting a demon inducing an abortion by pounding the abdomen of a pregnant woman with a pestle. (Source: Wikipedia: Abortion) |

The Guttmacher paper, "Estimating the Level of Abortion In the Philippines and Bangladesh" (1994), is the basis of the succeeding Guttmacher paper, The Incidence of Induced Abortion in the Philippines: Current Level and Recent Trends (2005). So I shall focus my analysis on the 1994 Guttmacher paper and show that the study could be improved by computing the error bars in each step of the methodology in order to arrive at the error bars for the estimates of abortion rates. The 1994 Guttmacher paper does not provide error bars in its estimates.

Williams Manual of Pregnancy Complications |

**A. Error bars in extrapolations**

Calculating the total number of hospitalized abortion patients. Of the 1,863 hospitals identified in the Philippines, we obtained usable reporting forms for 1,121.‡ We then made two basic adjustments to the data: If reporting forms were available for more than one year, the data were averaged; if the form covered only part of a year, the number of patients was adjusted to create an annual estimate, proportional to the number of months covered by the form.What we would like to see is a table showing the number of availability of data for each month per hospital for a given year. It is not clear what a "usable form" is. If there is only one month worth of data, the annual data would be estimated by multiplying by 12. But this assumes that the data is homogeneous per month, which may not be true because some months may peak and other months may be troughs. One possibility is to use hospitals with complete data sets as reference: the percentage contribution per month may be compared to see if some months really have peak values or not. The standard deviation of the percentage contributions per month may be computed and this would be used a reference for error estimates. Point blank estimates without error bars are deceptive, because we don't know how accurate these estimates are.

Hospital Operations: Principles of High Efficiency Health Care (FT Press Operations Management) |

**B. Error bars in percentage of abortions among the causes of hospitalizations**

For the remaining 776 hospitals, we assumed that admissions for abortion complications would account for about half as many patients as the number hospitalized for the lowest-ranking or the 10th-ranking cause.** This yielded an additional count of about 18,500 abortion complication cases per year from these 776 hospitals, for a combined total from the 1,121 hospitals with reporting forms of about 69,500 abortion-complication patients.

The third step was to estimate the likely annual number of abortion complications treated in the 742 hospitals for which there were no reporting forms. We used a regression equation in which the number of abortion patients was the dependent variable and hospital characteristics considered to be important determinants of the intake of abortion complication cases were analyzed. These characteristics were ownership (public vs. private), hospital level (primary, secondary or tertiary), hospital size (number of beds) and region. The regression equation was based on the 1,121 hospitals with information on the number of abortion patients, whether directly reported or estimated.Why insist on 1/2 of the number hospitalized for the lowest-ranking or 10th ranking cause? Why not 3/4 or 1/4 or 1/10? What we need is a table of leading 15 causes of hospital admissions per hospital and their percentages, with abortion complications at least covered. Is there a pattern in the ordering of the causes? What are the similar characteristics of hospitals with abortion complications among the top 10 of causes? What are the similar characteristics of hospitals with abortion outside the top 10 of causes? If we can identify these identifying characteristics for both cases, (e.g. public vs public hospitals), then we can make a reasonable guess on the the percentages of the abortion complication for hospitals that don't have existing data, by comparing it with a similar hospital in the same category that has statistics on abortion complications.

The authors may have used regression equation to estimate the number of abortion complications. But no equation was provided. What are the error bars? I think the only reasonable graph that can be done is hospital size vs number of abortion complications. A better parameter would be number of patients vs number of complications, so that we can make per capita estimates. Even if we use the number of hospital beds (hospital size) vs abortion complications per each of the six different category combinations (public vs public and primary vs secondary vs tertiary), and plot the data for each category combination, we still need to compute the error bars in the regression lines. One may expect that more hospital beds yields more abortion complications, but this may not be true, and the data would be scattered wildly that the errors in the estimation of abortion complications would increase. But we still need to state these errors, because it will affect the estimates for the induced abortion rate.

Avoiding Miscarriage: Everything You Need To Know To Feel More Confident In Pregnancy |

**C. Errors in proxy data sets for spontaneous miscarriage**

Because it has the advantage of being comparable across areas, an indirect method of estimating the number of women hospitalized for spontaneous miscarriages was used. In the absence of induced abortion, both the distribution of pregnancy loss by gestation and the proportion of live births among all pregnancies are fairly constant across populations. Such data are available both historically and from recent clinic-based studies in the United States and other countries.15Reference 15 is by Bongaarts and Potter which are also of the Guttmacher Institute. Is there another data set that is not connected whatsoever with Guttmacher Institute? The best data really is that of the same subjects in the hospitals under the study. The doctor asks each patient what medicine she took (e.g. abortion drug), then the doctor can declare whether the abortion complication was spontaneous or induced. But this data is not available. And if the RH law becomes operational, these types of questions may be forbidden, so that doctors may not pass moral judgments on the patients. Nevertheless, error bars must be provided for estimates for spontaneous abortions, because this will impact the induced abortion rate.

AHA Hospital Statistics Book and CD Combination, 2011 edition (Hospital Statistics (Book & CD-Rom)) |

**D. Conclusions and Recommendations**

**I limited this critique to three types of errors that may arise in the study. There are still other errors that I did not discuss, such as on the multipliers for the number of those with abortion complications who did not go to hospitals. If all these errors are taken into account, we can compute the upper and lower bounds of the estimate for induced abortion rate. But these error bars are not stated in this study.**

What DOH and the Philippine government can do is to provide a digital database of all hospitalizations in the country that any interested party may use to compute the abortion rates in order to verify and critique what other researchers have done. Simply parroting 11 mothers dying everyday is not scientific, and to base a law such as the RH law on flimsy data sets lacks prudence. Why do many Filipino mothers die every day? We need to know the real answers by careful isolation of the variables and parameters. Who knows, maybe the causes of maternal deaths are really hospital sanitation, incompetence of medical personnel, or neglect because of poverty--factors that are in no way related to induced abortion. We need to address these factors by placing systems and processes to diminish these causes, and thereby improve the quality of life of mothers.