SCM: a practical tool to implement hospital-based syndromic surveillance

Surveillance sentinels and targeted syndromes

Twenty-one hospitals were selected as sentinel sites for PD-SEWS, including two tertiary
hospitals, five secondary hospitals and fourteen primary hospitals. Hospitals were
chosen according to their location, catchment area, and patient volume 12]. In Pudong, a primary hospital is the smallest category of healthcare facility, providing
primary medical services such as medical treatment, prevention, healthcare, and rehabilitation
for a community with a population of 100,000 persons. The secondary hospitals provide
general medical services, including medical treatment, prevention, healthcare, and
rehabilitation for larger communities (population 100,000 persons). The tertiary
hospitals are regional healthcare centers providing specialized, high-complexity healthcare
services for several districts. Most of the sentinel sites were closed to the Exposition
venue, where was also the part of Pudong with highest population density (Fig. 1).

Fig. 1. The geographic location of sentinel hospitals at three different levels in Pudong
New Area, Shanghai, China, 2010

The concerning diseases during mass gatherings and their typical symptoms were listed
according to literature review, and Delphi method was employed to consult 18 domestic
epidemiologic and clinical professionals to score the disease severity, risk probabilities.
Then a disease-risk matrix was draw and 40 diseases were prioritized in the surveillance,
including local common diseases and some highly concerned diseases of importation.
The 25 most common and typical symptoms of the targeted diseases were enrolled and
classified to seven targeted syndromes: acute respiratory, acute gastrointestinal,
rash with fever, neurological syndrome, hemorrhagic fever, botulism-like syndrome
and acute viral hepatitis (Table 1).

Table 1. The seven syndromes under surveillance in the 41st exposition, Pudong New Area, Shanghai
City, China, 2010

Data collection and transmission

The SCM was developed and embedded in the HIS for each sentinel hospital, with the
25 symptoms presented on the SCM interface as a 5 × 5 table which could be selected
by single click on the screen (Fig. 2). In addition, we included an extra column to facilitate data entry and improve data
quality. If no symptoms in a row were presented, the ‘none of the left’ could be clicked.
A clinician could therefore check each row of the table quickly. For example, if a
patient described the chief complaint as ‘fever and cough’, the clinician should click
‘fever’ in the first row and ‘cough’ in the second, as well as ‘none of the left’
for the other three rows. Clinicians were prevented from moving to the next step of
treatment (such as prescribing a medication) until they finished recording symptoms
by clicking the ‘save’ button (Fig. 3).

Fig. 2. User interface of the symptom-clicking-module of PD-SEWS, Shanghai, China, 2010. All
English words in the SCM were translated from Chinese words

Fig. 3. Framework of syndromic surveillance in Pudong New Area, Shanghai, China, 2010

A record of symptoms in SCM was generated with a unique identity (ID) number and the
basic demographic information registered in HIS. All data were stored in real time
and transmitted automatically to the Pudong Public Health Database Center each day.
A virtual private network (VPN) was used to connect securely with sentinel hospitals,
and all data were supplied and analyzed in an anonymous format, without access to
personal identifying information. For improving the data quality, duplicated records
were rejected by checking the unique ID number, and if the data were not received
from a sentinel site, the Public Health Database Center would send a notice to that
hospital at 08:00 a.m. the next day. Once the database center received the surveillance
data, all of the patient records were automatically grouped and aggregated into the
seven syndromes according to the criteria listed in Table 1. If one patient’s symptoms referred to more than one syndrome, then the encounter
would be counted separately for each syndrome.

Data analysis and aberration detection

We developed an interface connecting to the Pudong Public Health Database Center.
The interface allowed authorized users to generate customized time series of total
visits for each syndrome stratified by gender, age, syndrome, hospital, start dates
and end dates. The cumulative sum (CUSUM) method, a control chart method commonly
used in syndromic surveillance 4], 13], was applied daily to analyze the aggregated data of all sentinel hospitals for detecting
abnormal temporal increases 14]. For each targeted syndrome, CUSUM compared the proportion of syndrome counts in
total visits in the current day (day 0) with the corresponding mean proportion and
standard deviation of the past 7 days (day-7 to day-1). A signal was generated if
the value of comparison exceeded two standard deviations.

The surveillance and response team of Pudong Center for Disease Control and prevention
(Pudong CDC) would monitor the warning signals routinely. When a signal was triggered,
the verification would be conducted immediately by analyzing and reviewing the data
to identify any unusual cluster of gender, age, occupation or hospital. Some signals
would be verified and compared with the data in the routine notifiable disease reporting
system, which recorded the confirmed patients’ detailed information. Potential epidemic
association would be explored by calling the clinicians or patients. If the signal
indicated a potential outbreak, field investigation would be performed to obtain more
detailed epidemiological information and necessary control measures would be conducted
to prevent further spread of the disease. Outbreak is defined as the occurrence of
cases of disease in excess of what would normally be expected in a defined geographical
area and period, which was further quantitatively defined for each kind of disease
in this study, so as to facilitate the outbreak confirmation and report. For example,
an outbreak of hand, foot and mouth (HFM) disease is defined as “within 1 week, at
least five HFM disease cases occur in the same setting-e.g. kindergarten or school-or
at least three cases of the disease occur in the same village or community” 15].

In this study, the number of outpatients, the reporting frequency and the proportion
of each syndrome were calculated. The average reporting frequencies of different syndromes
between weekdays and weekends were compared using a Chi square test, and a P value 0.05 was considered to be statistically significant.

Preparation for system operation

A series of training sessions on the system were conducted for the clinicians in the
sentinel hospitals before the practical operation. This approach helped to ensure
that the clinicians would be familiar with SCM in HIS. The estimated average time
cost of recording a patient’s syndrome by SCM was 10 s, according to our field investigation
in the last week of the training. A pilot surveillance operation was conducted 2 weeks
before the formal running of this syndromic surveillance system, and the staff from
Pudong CDC went to all the 21 sentinel hospitals during this pilot period to provide
help if needed.