Syndrome

As with the Poisson- and Bernoulli-based prospective space–time scan statistics [27 ], the space–time permutation scan statistic utilizes thousands or millions of overlapping cylinders to define the scanning window, each being a possible candidate for an outbreak. The circular base represents the geographical area of the potential outbreak. A typical approach is to first iterate over a finite number geographical grid points and then gradually increase the circle radius from zero to some maximum value defined by the user, iterating over the zip codes in the order in which they enter the circle. In this way, both small and large circles are considered, all of which overlap with many other circles. The height of the cylinder represents the number of days, with the requirement that the last day is always included together with a variable number of preceding days, up to some maximum defined by the user. For example, we may consider all cylinders with a height of 1, 2, 3, 4, 5, 6, or 7 d. For each center and radius of the circular cylinder base, the method iterates over all possible temporal cylinder lengths. This means that we will evaluate cylinders that are geographically large and temporally short, forming a flat disk, those that are geographically small and temporally long, forming a pole, and every other combination in between.
What is new with the space–time permutation scan statistic is the probability model. Since we do not have population-at-risk data, the expected must be calculated using only the cases. Suppose we have daily case counts for zip-code areas, where c_zd is the observed number of cases in zip-code area z during day d. The total number of observed cases (C) is

For each zip code and day, we calculate the expected number of cases μ_zd conditioning on the observed marginals:

In words, this is the proportion of all cases that occurred in zip-code area z times the total number of cases during day d. The expected number of cases μ_A in a particular cylinder A is the summation of these expectations over all the zip-code-days within that cylinder:

The underlying assumption when calculating these expected numbers is that the probability of a case being in zip-code area z, given that it was observed on day d, is the same for all days d.
Let c_A be the observed number of cases in the cylinder. Conditioned on the marginals, and when there is no space–time interaction, c_A is distributed according to the hypergeometric distribution with mean μ_A and probability function

When both Σ_zεAc_zd and Σ_dεAc_zd are small compared to C, c_A is approximately Poisson distributed with mean μ_A [37 ]. Based on this approximation, we use the Poisson generalized likelihood ratio (GLR) as a measure of the evidence that cylinder A contains an outbreak:

In words, this is the observed divided by the expected to the power of the observed inside the cylinder, multiplied by the observed divided by the expected to the power of the observed outside the cylinder. Among the many cylinders evaluated, the one with the maximum GLR constitutes the space–time cluster of cases that is least likely to be a chance occurrence and, hence, is the primary candidate for a true outbreak. One reason for using the Poisson approximation is that it is much easier to work with this distribution than the hypergeometric when adjusting for space by day-of-week interaction (see below), as the sum of Poisson distributions is still a Poisson distribution.
Since we are evaluating a huge number of outbreak locations, sizes, and time lengths, there is serious multiple testing that we need to adjust for. Since we do not have population-at-risk data, this cannot be done in any of the usual ways for scan statistics. Instead, it is done by creating a large number of random permutations of the spatial and temporal attributes of each case in the dataset. That is, we shuffle the dates/times and assign them to the original set of case locations, ensuring that both the spatial and temporal marginals are unchanged. After that, the most likely cluster is calculated for each simulated dataset in exactly the same way as for the real data. Statistical significance is evaluated using Monte Carlo hypothesis testing [38 ]. If, for example, the maximum GLR is calculated from 999 simulated datasets, and the maximum GLR for the real data is higher than the 50th highest, then that cluster is statistically significant at the 0.05 level. In general terms, the p-value is p = R/(S + 1) where R is the rank of the maximum GLR from the real dataset and S is the number of simulated datasets [38 ]. In addition to p-values, we also report null occurrence rates [8 (link)], such as once every 45 d or once every 23 mo. The null occurrence rate is the expected time between seeing an outbreak signal with an equal or higher GLR assuming that the null hypothesis is true. For daily analyses, it is defined as once every 1/p d. For example, under the null hypothesis we would at the 0.05 level on average expect one false alarm every 20 d for each syndrome under surveillance.
Because of the Monte Carlo hypothesis testing, the method is computer intensive. To facilitate the use of the methods by local, state, and federal health departments, the space–time permutation scan statistic has been implemented as a feature in the free and public domain SaTScan software [36 ].

We performed a focused literature search and consulted with a working group of experienced clinicians, researchers, and patient partners to collate a list of frailty-related terms. This approach was adapted from Urquhart et al.’s development of a rule to identify frailty in administrative health databases [25 (link)]. The aim was to construct a list of terms or phrases that would commonly be used by providers to describe patients living with frailty.
The focused literature review involved an initial scan of PubMed using combinations of the following terms: “frail elderly”, “frailty”, “identification”, “definition”, “database”, and “health data”. We were particularly interested in studies that have used key-term searching to identify frailty in the free text of other healthcare datasets. We considered a variety of study types, including systematic reviews and other evidence syntheses, clinical guidelines, retrospective studies of healthcare or administrative databases, and studies that have developed or validated frailty assessment tools. Search results were supplemented by articles recommended by the research team and a hand search of reference lists of selected articles.
Relevant findings from the literature search were summarized. From each included study, we extracted terms related to the identification or assessment of frailty, including but not limited to signs and symptoms, comorbidities, disabilities, and related clinical syndromes.
To ensure the content validity of our selection, the preliminary list of frailty-related terms was then shared for feedback with a working group of clinicians (n = 4), researchers (n = 4) and a patient partner (n = 1) who are knowledgeable about LTC, primary care and frailty, each bringing diverse perspectives on these topics (see corresponding section in Supplemental Methods). Through iterative discussions and revisions, a version of the list of frailty terms was finalized for a key-term search of the eConsult text. The list was organized by grouping terms into overarching topic categories.

We collected basic information, including age, gender, education, onset, and duration of MDD, and married status through a self-designed questionnaire. All participants were independently interviewed face-to-face by two trained psychiatrists via the SCID. Two psychiatrists independently assessed each participant’s depression, anxiety, and psychotic symptoms by the HAMD, Hamilton Anxiety Scale (HAMA), and the positive subscale of Positive and Negative Syndrome Subscale (PANSS), respectively. HAMD score ranges from 0-52, with a cutoff point of 24 being used to determine severe depression (28 (link)). HAMA consists of 14 items, measuring psychological and somatic anxiety symptoms (29 (link)). It applied the 5-Likert scale, with a total score ranging from 0-56. The PANSS positive subscale assesses seven positive symptoms (30 (link)). The PANSS-positive subscale score ranges from 7-49. Higher scores on the HAMA, HAMD, and PANSS indicate more severe symptoms. These three scales have been validated and widely used in the Chinese population (31 (link)–33 (link)). According to previous studies (34 (link), 35 (link)), HAMA score >20 and PANSS positive subscale score >14 indicate significant anxiety and psychotic symptoms, respectively. The correlation coefficients between the two psychiatrists’ scores on all three scales were higher than 0.8.
We assessed SA through face-face interviews. All participants were asked the question: “In your lifetime, did you ever try to kill yourself?”. This single item has been validated and used widely in previous epidemiological studies for the detection of SA (36 (link), 37 (link)). Those who answered “yes” were considered to have lifetime SA. We further asked them about the timing and frequency of SA. We contacted the family members of the participants for the details of SA when patients were unable to provide definitive information.