Informatics as Support for Changes in Health Policy: A Case in

Obstetrics

Giovana Jaskulski Gelatti

1,2 a

, Pedro Pereira Rodrigues

2,3 b

and Ricardo Jo

ao Cruz Correia

2,3 c

Institute of Mathematical and Computer Sciences of University of S

ao Paulo, 13566-590, S

ao Carlos, Brazil

Centre for Health Technology and Services Research, 4200-450, Porto, Portugal

Universidade do Porto, Faculty of Medicine of University of Porto, 4200-319, Porto, Portugal

Keywords:

Health Policies, Medical Information, Health Data Science, Computer Support, Robson Classiﬁcation,

Obstetrics.

Abstract:

Introduction: In 2015 the Directorate-General for Health of Portugal published new standards (DGS

001/2015) for the registration of cesarean section indicators. The existing scenario was the lack of data,

inﬂuencing the quality of indicators and analyses on them. The use of a single computer tool was encouraged

to register and compare indicators between hospitals with special attention to the Robson Classiﬁcation as it

employs basic information of pregnancy to classify all deliveries in 10 groups. The selected tool was Obscare

software.

Aim: Describe the scenario on data quality by analyzing the completeness of obstetric records from 2016 to

2018 of the variables used in Robson’s classiﬁcation collected by the Obscare tool.

Methods: The completeness is evaluated using a number of missing values. The lower the completeness, the

higher the number of missing values. Also, we perform the imputation of data based on basic concepts and

analyzed the participation of this data in the indication of the type of delivery to be performed according to

classiﬁcation suggested by DGS 001/2015.

Results: From 2016 to 2018, 5922 number of pregnancies resulted in 5922 of Robson Classiﬁcations. The

variables with lower completeness were related to previous cesarean section (77%) and previous pregnancies

(43%). After imputation, it fell to 3.9% and 0.56%, respectively causing 4.6% of discarded data from the total.

Discussion: There is a signiﬁcant amount of missing data in basic variables used to study the classiﬁcation of

delivery type. We believe that encouraging data completion with the possibility of comparing data between

hospitals should be a priority in the health area.

1 INTRODUCTION

Cesarean section is a surgery that provides a high risk

of complications to both the pregnant woman and the

baby, its rate is used as one of the health indicators.

However, the current scenario reveals high rates of

this type of delivery (Betr

an et al., 2016) and the de-

crease in cesarean rate has become a governmental

concern. To reverse this framework, changes in health

policies have been designed.

In Portugal, the Directorate-General for Health

(DGS, Portuguese acronym) published Standard No.

001/2015 which deﬁnes concepts of cesarean sec-

https://orcid.org/0000-0003-1003-7010

https://orcid.org/0000-0001-7867-6682

https://orcid.org/0000-0002-3764-5158

tion, reasons for its performance and classiﬁcations

of cesarean types regarding the urgency of surgery,

absence or labour phase of delivery, main reason

of indication to surgery and main characteristics of

pregnancy, indicating the Classiﬁcation of Robson

(Robson, 2001) for the latter. By implementing na-

tional health policy in northern Portugal, the Regional

Health Administration of the North I.P. (ARSN), re-

sponsible for propose the committee for the reduc-

tion of the rate of cesarean sections (CRTC). This

committee, when evaluating the frequency and rea-

sons for cesarean sections, was faced with the lack of

standardized computerization of maternal-fetal data,

which makes it difﬁcult to understand the current sit-

uation. As a result, one of the ﬁrst measures of the

CRTC was to propose ”the implementation of a single

computer program for hospital registration of perina-

Gelatti, G., Rodrigues, P. and Correia, R.

Informatics as Support for Changes in Health Policy: A Case in Obstetrics.

DOI: 10.5220/0009173807450749

In Proceedings of the 13th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2020) - Volume 5: HEALTHINF, pages 745-749

ISBN: 978-989-758-398-8; ISSN: 2184-4305

 2022 by SCITEPRESS – Science and Technology Publications, Lda. All rights reser ved

745

tal data (...) to allow the automatic export of perinatal

indicators to ARSN” and ”annual public disclosure by

the ARNS of the cesarean rate of each hospital cen-

ter”(Campos et al., 2010), thus contributing to the im-

provement of indicators.

One system used in hospitals in Portugal that

meets all DGS and ARSN requirements and chosen

by the Northern Region Operational Programme for

this purpose is Obscare. Obscare is used in 15 hospi-

tals in Portugal and through Obscare, the reports com-

posed of obstetric indices and all required in the DGS

Standard are generated.

This article aims to describe the current scenario

on data quality after changes required by DGS and

ARSN. Besides, some ﬁndings and discussions are re-

ported.

1.1 Implementation of the Policies

Required by DGS

In 2015, the carelessness was reported with health

data and, for this reason, hospitals are required to fol-

low Standard No. 001/2015. The standard requires

hospitals to deliver to DGS a detailed annual report

with the deﬁnitions described in its standard. One of

the deﬁnitions is to include Robson Classiﬁcation on

these reports.

The Robson Ten Group classiﬁcation (RTGC) has

four criteria and, from them, all pregnant women

can be classiﬁed into 10 mutually exclusive and fully

inclusive groups. The criteria are on obstetric his-

tory (nulliparous or multiparous with or without an-

terior cesarean section), the type of pregnancy (sin-

gle cephalic or pelvic or transverse fetus, multiple

pregnancy), how childbirth is triggered (spontaneous,

induced or cesarean section and the gestational age

at which childbirth occurs (before or from the 37th

week). Based on these criteria, pregnant women are

classiﬁed into one of the groups described in Table 1.

The criteria are of simple identiﬁcation and classiﬁ-

cation, clinically relevant. And so, obstetric centres

from around the world begin to use this classiﬁcation.

1.2 Obscare

Obscare (see http://virtualcare.pt/portfolio/vc-

obscare/) is an electronic clinical obstetrics and

gynaecology registration system designed to be used

by doctors, anesthesiologists, nurses and adminis-

trative staff. The system is installed in 15 hospitals

in Portugal, 37% of Portugal’s public hospitals

(data provided by Virtualcare enterprise, owner of

Obscare), and has a proper semi-structured form with

support for user input per clinical event. Each event

Table 1: Characteristics of each group of Robson’s classiﬁ-

cation.

Group 1 Nulliparous, singleton, cephalic,

≥37 weeks’ gestation, in sponta-

neous labour

Group 2 Nulliparous, singleton, cephalic,

≥37 weeks’ gestation, induced

labour or caesarean section be-

fore labour

Group 3 Multiparous (excluding previous

caesarean section), singleton,

cephalic, ≥37 weeks’ gestation,

in spontaneous labour

Group 4 Multiparous without a previ-

ous uterine scar, with singleton,

cephalic pregnancy, ≥37 weeks’

gestation, induced or caesarean

section before labour

Group 5 Previous caesarean section, sin-

gleton, cephalic, ≥37 weeks’

gestation

Group 6 All nulliparous with a single

breech pregnancy

Group 7 All multiparous with a single

breech (including previous cae-

sarean section)

Group 8 All multiple pregnancies (includ-

ing previous caesarean section)

Group 9 All women with a single preg-

nancy in transverse or oblique

lie (including those with previous

caesarean section)

Group 10 All singleton, cephalic, <37

weeks’ gestation pregnancies (in-

cluding previous caesarean sec-

tion)

has forms deﬁned with experts in the ﬁeld that meet

all requirements deﬁned in DGS standards. It is also

possible to export monthly, quarterly, half-yearly

and annual reports used to compare the indicators

and data quality in these periods. In addition to the

presence in several hospitals in Portugal, there was a

growing use of the system and records for scientiﬁc

production in the country (Pereira et al., 2019). This

growing demand for the use of its data has generated

concern about the quality of what is being shared.

Obscare discloses in its reports the information on

the quality of data in the period examined. Informing

the quality of data to hospitals encourages the search

for an increasing data ﬁll. Although it is a robust

and complete system that allows the registration and

analysis of data for hospitals, the quality of health

research is related to this quality in ﬁlling out data by

users.

2 METHODS

Despite the concern stemming from data computeri-

zation and implementation of new policies, many ret-

rospective data are uncertain. When deploying the

HEALTHINF 2020 - 13th International Conference on Health Informatics

746

new rules, cases of empty of basic data for classiﬁ-

cation or any use of the data were reported. (Vogel

et al., 2015) says that ”the population level CS rate is

too crude to be useful as clinical indications for CS

data are missing” and this causes a great risk to the

academic community that relies on the data for re-

production and development of new knowledge from

them.

From 2015 to the current year, hospitals paid

greater attention to the misconceptions of ﬁlling and

validating their data to ﬁt DGS standards. Hospitals

were required to adopt the rules for building DGS re-

ports, including Robson’s classiﬁcation and interpret-

ing the classiﬁcation for an improvement in their data

and reporting cesarean rates in their hospitals.Despite

the new rules, there are missing data in the vari-

ables used for RTGC (Begum et al., 2019; Linard

et al., 2019; Senanayake et al., 2019; Ming et al.,

2019; Zimmo et al., 2018; Kacerauskiene et al., 2018)

and limitations in the application of the same (Betr

et al., 2014; Rebelo et al., 2010). As small as the limi-

tations for applying classiﬁcation or missing data are,

the variables used are basic for obstetric data analysis.

To develop the analysis on the quality of health

data in recent years, data on obstetrics, speciﬁcally the

variables used for Robson’s classiﬁcation of a hospital

using the Obscare system, were used. An analysis of

data completeness and deterministic imputation was

performed in missing data.

2.1 Case of Study

The data collected by the Obscare system of a Por-

tuguese hospital refer to the variables used for Rob-

son’s classiﬁcation in 2016, 2017 and 2018. At ﬁrst

instance, a large number of missing data was found.

However, it was possible to perform a deterministic

imputation through the concepts present in non-empty

variables, for example, as per Robson’s associated

group. With most of the records identiﬁed by these

groups, it was possible to ﬁll values on the character-

istics associated with the group, performing a deter-

ministic imputation.

Another deﬁnition was the variable of the number

of deliveries prior to pregnancy. Observations classi-

ﬁed as group 1, 2 or 6 are nulliparous, without pre-

vious deliveries, so the veriﬁcation and/or imputation

of value 0 was applied in the delivery variable already

made. It also serves in the case of the variable regard-

ing the number of pregnancies, including current: if

1 is the ﬁrst pregnancy and there are no pregnancy or

background deliveries, the amount of pregnancies and

previous deliveries is 0.

The observations classiﬁed in group 1 and 2 of

Robson’s classiﬁcation refer to nulliparous pregnant

women and those in group 3 and 4 are multiparous

but without antecedent cesarean sections. The vari-

able referring to previous cesarean sections of the ob-

servations identiﬁed in these groups were ﬁlled with

the zero value.

Information was also searched in structured or

free text variables that could have relevant informa-

tion from each observation with the unknown Robson

group for it to be ﬁlled. In total, there were 39 ob-

servations without group identiﬁcation. Using vari-

ables of cesarean section motif and diagnoses that

contained obstetric antecedents, Robson’s group of 4

observations were identiﬁed.

13 cases with gestation week values in -1 were an-

alyzed on a case-by-case basis. 12 cases were classi-

ﬁed in group 10 corresponding to pregnant women of

a baby with cephalic presentation and preterm. The

value was corrected for the average of weeks of ges-

tation in group 10 (35 weeks). No relevant informa-

tion was found for ﬁlling weeks of gestation from the

remaining observation since it is classiﬁed as group 6

and this does not concern this. For this reason, it was

ﬁlled with the average of weeks of gestation (

38.94,

was completed as 39 weeks).

3 RESULTS

Performing a simple deterministic imputation, we

were able to decrease the missing rate from 42.6 to

0.56%, ﬁlling out the observations regarding the data

of previous pregnancy, and from 76.97 to 3.91% in

observations ﬁlled with data on Previous Cesarean

sections. After completing the data through deter-

ministic imputation, 5922 valid records were selected,

representing 95.34% of the total observations (4.66%

of missing values).

Table 2: Table referring to the amount of missing data on the

total observations before and after deterministic imputation.

Variable

Missings (%total)

data after imputation

Robson’s group 39 (0,6279) 35 (0,5635)

Previous pregnancies 2646 (42,6018) 35 (0,5635)

Previous cesarean sections 4781 (76,9763) 243 (3,9124)

The contribution of each variable to cesarean

delivery, including Robson’s associated group, de-

scribed in Table 3, also present another form of inter-

pretation of Robson’s groups as it indicated. In it, the

groups are also classiﬁed as very preventable cesarean

section (groups 1, 2, 3 and 4), avoidable (5, 6 and 9) or

not avoidable (7, 8 and 10), indicated in the interpre-

Informatics as Support for Changes in Health Policy: A Case in Obstetrics

747

tation of RTGC. The probability of signiﬁcance, p, re-

sulting from statistical tests, Pearson or Fisher, is ag-

gregated. The χ

, or Pearson test was used in samples

with frequencies greater than 20% of the expected fre-

quency in observations per case in contingency tables.

This is thus deﬁned because in small samples the ap-

proximation of the χ distribution is affected and not

considered satisfactory. For the other cases, the exact

Fisher test was used (Christensen, 2005).

Table 3: Frequencies and p value produced.

Values

Cesarean

p − value

Total

n (%) n (%)

Total 1316 (22.22) 5922 (100)

Number of fetuses 1 1191 (21.31)

<0.001

5589 (98.97)

2 or more 40 (68.97) 58 (1.03)

Robson Groups CS very preventable 533 (11.92) <0.001 4471 (79.17)

Group 1 159 (9.61) 1654 (29.29)

Group 2 281 (36.07) 779 (13.79)

Group 3 37 (2.30) 1608 (28.48)

Group 4 56 (13.02) 430 (7.61)

CS avoidable 549 (65.12) 843 (14.93)

Group 5 406 (58.17) 698 (12.36)

Group 6 143 (98.62) 145 (2.57)

Group 9 0 (0) 0 (0)

CS not preventable 149 (44.74) 333 (5.90)

Group 7 65 (100) 65 (1.15)

Group 8 40 (67.80) 59 (1.04)

Group 10 44 (21.05) 209 (3.70)

Labour absent 484 (99.38) 0 487 (8.62)

induced 330 (25.80) 1279 (22.65)

spontaneous 417 (10.74) 3881 (68.73)

Previous deliveries

0 683 (23.57)

0.04

2898 (48.94)

1 482 (20.65) 2334 (39.41)

2 or more 151 (21.88) 690 (11.65)

Pregnancies 1 560 (22.54)

0.15

2485 (41.96)

2 454 (20.94) 2168 (36.61)

3 191 (22.82) 837 (14.13)

4 or more 111 (25.69) 432 (7.29)

Previous CS

0 806 (15.89)

<0.001

5073 (85.66)

1 405 (54.51) 743 (12.55)

2 or more 105 (99.06) 106 (1.79)

* CS = cesarean section

In Table 3, it is seen that the highest cesarean

section is of the preventable cesarean section group

(65.12% of cases). This group is composed of groups

5, 6 and 9, the latter of which had no contribution.

4 DISCUSSION

The largest amount of data taken from the analysis

was made due to the missing data on the cesarean

section before pregnancy. Although most excluded

cases had this characteristic, having previous cesarean

section the pregnancy was also the characteristic of

the group with the greatest contribution to the per-

formance of cesarean delivery. According to RTGC,

65% of deliveries considered cesarean sections avoid-

able belong to pregnant women with a history of ce-

sarean section or pregnant women with breech preg-

nancy with no history of pregnancies. The great con-

tribution of cesarean deliveries in pregnant women

who already have a history with this type of delivery,

suggests that greater information should be made and

implementation of discussions about the dangers that

cesarean section causes the pregnant woman and the

baby.

Furthermore, may the classiﬁcation does not re-

ﬂect the real reasons for a cesarean section to have

been performed. Therefore, it cannot be afﬁrmed that

the result of the classiﬁcation in this group is faith-

ful to the condition that the delivery was performed.

The inclusion of other variables or alteration of the

variables that the classiﬁcation uses results in a more

accurate description, as well as in the reinforcement

in the completion of the data.

The high rate of cesarean sections classiﬁed as

preventable and the lack of completeness and data

sharing between hospitals remains worrying and may

be related. We believe that measures to cohere and en-

courage data completion with the possibility of com-

paring the rates among hospitals should be a main

concern and priority in the health area. With increased

information sharing between hospitals, the exchange

of health strategies increases, which can generate a

global improvement in health and data quality indices.

To improve it quality and analysis, simple strategies

can be adopted such as deterministic imputation en-

sured the reduction of missing data; which shows the

lack of completion of these by hospital units.

Other computer techniques for imputation and dis-

covery of new knowledge, as proposed by the ap-

proach of machine learning and data mining tech-

niques, may be another alternative for increasing data

quality. Their use in the clinical, medical and data

management context can help routine and discover

strategies for reducing cesarean rate, for example. As

future work, the study of computational techniques

used in the obstetric context for imputation or a new

discovery in the area will be developed, besides con-

tinuing the study and analysis on the signiﬁcance of

obstetric variables in the classiﬁcation of the type of

delivery.

5 CONCLUSION AND FUTURE

WORK

Health policy change has encouraged the creation of

tools for storing and visualizing data. With the imple-

mentation of tools, the analysis of data and especially

the quality of data created is made easier. Quality is

directly affected by the number of ﬁelds ﬁlled in by

professionals and by the strategies adopted in storage.

It can be seen that simple auto complete strategies can

be adopted to increase quality. This study showed that

identifying simple strategies with deterministic impu-

tation ensured the reduction of missing data. On this

case, only 4.66% of the data was discarded.

HEALTHINF 2020 - 13th International Conference on Health Informatics

748

To add, our study showed the contribution of each

group of Robson in the cesarean section and that the

group with the highest index is deﬁned as preventable

cesarean sections. We also demonstrated that the vari-

ables used for this classiﬁcation are signiﬁcant to the

analysis. This importance is reﬂected in the addi-

tion of features that inform the quality per registra-

tion tool. However, there is a signiﬁcant amount of

missing data in basic variables used to study the clas-

siﬁcation of delivery type. We believe that encourag-

ing data completion with the possibility of compar-

ing data between hospitals should be a priority in the

health area.

REFERENCES

(2015). Norma 001/2015: Indicac¸

oes de cesariana.

Direc¸

ao-Geral da Sa

ude. Minist

erio da Sa

ude. Lis-

boa.

Begum, T., Nababan, H., Rahman, A., Islam, M. R., Adams,

A., and Anwar, I. (2019). Monitoring caesarean births

using the robson ten group classiﬁcation system: A

cross-sectional survey of private for-proﬁt facilities

in urban bangladesh. PloS one, 14(8):e0220693–

e0220693.

Betr

an, A. P., Vindevoghel, N., Souza, J. P., G

ulmezoglu,

A. M., and Torloni, M. R. (2014). A systematic re-

view of the robson classiﬁcation for caesarean section:

what works, doesn’t work and how to improve it. PloS

one, 9(6):e97769.

Betr

an, A. P., Ye, J., Moller, A.-B., Zhang, J., G

ulmezoglu,

A. M., and Torloni, M. R. (2016). The increas-

ing trend in caesarean section rates: global, re-

gional and national estimates: 1990-2014. PloS one,

11(2):e0148343.

Campos, D., Furtado, J., Cris

ostomo, M., Carrapato, R.,

Cunha, E., and Conceic¸

ao, M. (2010). Medidas para

reduzir a taxa de cesarianas na regi

ao norte de portu-

gal. Portugal: Comiss

ao para a reduc¸

ao da taxa de

cesarianas da ARS Norte.

Christensen, R. (2005). Testing ﬁsher, neyman, pearson,

and bayes. The American Statistician, 59(2):121–126.

Kacerauskiene, J., Minkauskiene, M., Mahmood, T., Bartu-

seviciene, E., Railaite, D. R., Bartusevicius, A., Kli-

ucinskas, M., Nadisauskiene, R. J., Smigelskas, K.,

Maciuliene, K., et al. (2018). Lithuania’s experi-

ence in reducing caesarean sections among nulliparas.

BMC pregnancy and childbirth, 18(1):419.

Linard, M., Deneux-Tharaux, C., Luton, D., Schmitz, T.,

Mandelbrot, L., Estellat, C., Sauvegrain, P., Azria, E.,

study group, P., study group, B., et al. (2019). Dif-

ferential rates of cesarean delivery by maternal geo-

graphical origin: a cohort study in france. BMC preg-

nancy and childbirth, 19(1):217.

Ming, Y., Li, M., Dai, F., Huang, R., Zhang, J., Zhang, L.,

Qin, M., Zhu, L., Yu, H., and Zhang, J. (2019). Dis-

secting the current caesarean section rate in shanghai,

china. Scientiﬁc reports, 9(1):2080.

Pereira, A. M., Maranh

ao, P. A., Silva-Costa, T., and Cor-

reia, R. J. C. (2019). The impact of electronic ob-

stetric records using the obscare software on scientiﬁc

production. In EFMI-STC, pages 55–59.

Rebelo, F., Da Rocha, C. M. M., Cortes, T. R., Dutra, C. L.,

and Kac, G. (2010). High cesarean prevalence in a na-

tional population-based study in brazil: the role of pri-

vate practice. Acta obstetricia et gynecologica Scan-

dinavica, 89(7):903–908.

Robson, M. S. (2001). Classiﬁcation of caesarean sections.

Fetal and maternal medicine review, 12(1):23–39.

Senanayake, H., Piccoli, M., Valente, E. P., Businelli, C.,

Mohamed, R., Fernando, R., Sakalasuriya, A., Ih-

san, F. R., Covi, B., Wanzira, H., et al. (2019). Im-

plementation of the who manual for robson classi-

ﬁcation: an example from sri lanka using a local

database for developing quality improvement recom-

mendations. BMJ open, 9(2):bmjopen–2018.

Vogel, J. P., Betr

an, A. P., Vindevoghel, N., Souza, J. P.,

Torloni, M. R., Zhang, J., Tunc¸alp,

O., Mori, R.,

Morisaki, N., Ortiz-Panozo, E., et al. (2015). Use of

the robson classiﬁcation to assess caesarean section

trends in 21 countries: a secondary analysis of two

who multicountry surveys. The Lancet Global Health,

3(5):e260–e270.

Zimmo, M. W., Laine, K., Hassan, S., Bottcher, B., Fosse,

E., Ali-Masri, H., Zimmo, K., Falk, R. S., Lieng, M.,

and Vikanes,

A. (2018). Caesarean section in pales-

tine using the robson ten group classiﬁcation system:

a population-based birth cohort study. BMJ open,

8(10):e022875.

Informatics as Support for Changes in Health Policy: A Case in Obstetrics

749