R中死亡率指标的7天滚动平均值答案

【问题标题】：7 day rolling average of mortality rate metric in RR中死亡率指标的7天滚动平均值
【发布时间】：2021-03-24 18:54:54
【问题描述】：

我有一个医院患者的数据集，其中包含他们的入院日期和结果（死亡或出院）。 对于每一天，我想计算死亡率的 7 天滚动平均值。具体来说，我希望代码 a) 添加没有患者入院的缺失日期（直到现在我已经手动完成了 --> 是在此处创建第二个数据集和 left_join() 的最佳选择吗？） b) 对于每一天，确定过去 7 天（即最后 6 天加上手头的具体日期）内入院的所有患者 c) 根据这 7 天窗口内入院患者的结果（=（死亡患者数 / 所有患者数）计算死亡率

示例：如果我查看 4 月 8 日，我想知道这天和前 6 天入院的患者有哪些。然后我想让代码知道样本中有 x 个死亡患者和 y 个出院患者，并根据 x/x+y 计算死亡率。

之前我将患者分配到第 1 周、第 2 周、第 3 周等，并使用 dplyr 的 group_by() 和 summarise() 来计算每周的死亡率。现在，我必须做一个滚动平均，但我不知道该怎么做。我已经使用 R 有一段时间了，但有时仍然觉得自己像个初学者 ://

这是一些数据：

structure(list(Summary = c("Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Dead", "Discharged", "Discharged", 
"Discharged", "Dead", "Dead", "Dead", "Dead", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Dead", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Dead", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Dead", "Discharged", 
"Dead", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Dead", "Discharged", "Discharged", "Dead", "Discharged", 
"Discharged", "Discharged", "Dead", "Dead", "Discharged", "Discharged", 
"Dead", "Discharged", "Dead", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Dead", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Dead", "Discharged", 
"Dead", "Discharged", "Discharged", "Dead", "Discharged", "Discharged", 
"Dead", "Dead", "Discharged", "Discharged", "Discharged", "Dead", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Dead", "Discharged", "Discharged", "Dead", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Dead", "Discharged", 
"Dead", "Discharged", "Dead", "Dead", "Dead", "Discharged", "Discharged", 
"Discharged", "Dead", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Dead", "Discharged", "Dead", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Dead", "Dead", "Discharged", "Discharged", 
"Discharged", "Dead", "Discharged", "Dead", "Discharged", "Discharged", 
"Dead", "Discharged", "Dead", "Discharged", "Discharged", "Dead", 
"Discharged", "Discharged", "Discharged", "Dead", "Discharged", 
"Dead", "Discharged", "Discharged", "Discharged", "Dead", "Discharged", 
"Discharged", "Discharged", "Dead", "Discharged", "Dead", "Dead", 
"Discharged", "Dead", "Dead", "Discharged", "Discharged", "Dead", 
"Dead", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Dead", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Dead", "Discharged", "Dead", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Dead", "Discharged", 
"Dead", "Discharged", "Dead", "Discharged", "Discharged", "Discharged", 
"Dead", "Dead", "Dead", "Discharged", "Discharged", "Dead", "Dead", 
"Discharged", "Discharged", "Discharged", "Dead", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Dead", "Dead", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Dead", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Dead", "Discharged", 
"Discharged", "Dead", "Discharged", "Dead", "Discharged", "Dead", 
"Dead", "Discharged", "Discharged", "Discharged", "Dead", "Dead", 
"Discharged", "Dead", "Dead", "Discharged", "Dead", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Dead", 
"Discharged", "Discharged", "Discharged", "Discharged", "Dead", 
"Dead", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Dead", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Dead", "Discharged", "Discharged", 
"Discharged", "Discharged", "Dead", "Dead", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Dead", "Discharged", 
"Dead", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Dead", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Dead", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Dead", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Dead", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Dead", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Dead", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged", "Discharged", 
"Discharged", "Discharged", "Dead", "Discharged", "Discharged", 
"Discharged", "Discharged", "Discharged", "Discharged"), Admission = structure(c(1578586420.224, 
1580147094.528, 1580467041.28, 1580482376.704, 1581576565.76, 
1582319350.784, 1582387115.008, 1582472442.88, 1582575203.328, 
1582861464.576, 1583030023.168, 1583160439.808, 1583170532.352, 
1583178658.816, 1583234757.632, 1583471735.808, 1583610278.912, 
1583645930.496, 1583763633.152, 1583833887.744, 1583926817.792, 
1583934813.184, 1583949231.104, 1584037966.848, 1584093803.52, 
1584104289.28, 1584196039.68, 1584211506.176, 1584287003.648, 
1584290280.448, 1584340218.88, 1584449139.712, 1584453989.376, 
1584522540.032, 1584586503.168, 1584596071.424, 1584615994.368, 
1584620581.888, 1584627266.56, 1584636703.744, 1584643388.416, 
1584646140.928, 1584688477.184, 1584701846.528, 1584715215.872, 
1584737104.896, 1584792941.568, 1584803165.184, 1584817976.32, 
1584874075.136, 1584892949.504, 1584953111.552, 1584959403.008, 
1584962417.664, 1584966480.896, 1584990336, 1584991253.504, 1585006195.712, 
1585032147.968, 1585044599.808, 1585052595.2, 1585057707.008, 
1585059804.16, 1585066488.832, 1585080775.68, 1585082217.472, 
1585105286.144, 1585108562.944, 1585133990.912, 1585140151.296, 
1585142903.808, 1585142903.808, 1585150505.984, 1585161122.816, 
1585168331.776, 1585171084.288, 1585219843.072, 1585237406.72, 
1585245533.184, 1585254315.008, 1585267553.28, 1585269781.504, 
1585279742.976, 1585301107.712, 1585376736.256, 1585391285.248, 
1585393120.256, 1585396134.912, 1585397445.632, 1585399542.784, 
1585412387.84, 1585413436.416, 1585422349.312, 1585427592.192, 
1585432966.144, 1585435456.512, 1585464292.352, 1585485394.944, 
1585505711.104, 1585509512.192, 1585516065.792, 1585518162.944, 
1585524061.184, 1585567183.872, 1585567314.944, 1585573999.616, 
1585578456.064, 1585611355.136, 1585615811.584, 1585641763.84, 
1585643860.992, 1585646482.432, 1585660507.136, 1585684886.528, 
1585686983.68, 1585689605.12, 1585729844.224, 1585737446.4, 1585740461.056, 
1585748718.592, 1585751995.392, 1585776374.784, 1585778603.008, 
1585780962.304, 1585788302.336, 1585819759.616, 1585836667.904, 
1585852920.832, 1585863144.448, 1585873761.28, 1585930122.24, 
1585931301.888, 1585942836.224, 1585956336.64, 1585997231.104, 
1586003522.56, 1586007061.504, 1586008503.296, 1586018989.056, 
1586020561.92, 1586092913.664, 1586094486.528, 1586094879.744, 
1586095928.32, 1586099598.336, 1586107200.512, 1586107855.872, 
1586109821.952, 1586109821.952, 1586121225.216, 1586162643.968, 
1586171032.576, 1586182042.624, 1586182304.768, 1586185188.352, 
1586193052.672, 1586198819.84, 1586205897.728, 1586207732.736, 
1586264486.912, 1586268812.288, 1586273399.808, 1586273924.096, 
1586274579.456, 1586275234.816, 1586281264.128, 1586288342.016, 
1586290308.096, 1586293715.968, 1586337100.8, 1586339460.096, 
1586343392.256, 1586366854.144, 1586370393.088, 1586376422.4, 
1586379043.84, 1586379043.84, 1586384024.576, 1586393986.048, 
1586426229.76, 1586433438.72, 1586452050.944, 1586465944.576, 
1586473808.896, 1586513523.712, 1586517980.16, 1586523091.968, 
1586527417.344, 1586531742.72, 1586542359.552, 1586544718.848, 
1586555073.536, 1586595443.712, 1586597671.936, 1586609337.344, 
1586614973.44, 1586617594.88, 1586630702.08, 1586636993.536, 
1586670941.184, 1586681558.016, 1586685883.392, 1586703709.184, 
1586711966.72, 1586722976.768, 1586776978.432, 1586779993.088, 
1586886423.552, 1586891797.504, 1586897302.528, 1586924696.576, 
1586924958.72, 1586937672.704, 1586973062.144, 1586982892.544, 
1586985251.84, 1586988397.568, 1587024180.224, 1587024573.44, 
1587046855.68, 1587075167.232, 1587135198.208, 1587137164.288, 
1587138081.792, 1587142538.24, 1587146994.688, 1587169276.928, 
1587203879.936, 1587232060.416, 1587255129.088, 1587257619.456, 
1587266925.568, 1587277149.184, 1587283309.568, 1587305722.88, 
1587308475.392, 1587330757.632, 1587377419.264, 1587379385.344, 
1587383448.576, 1587397080.064, 1587409925.12, 1587413988.352, 
1587425391.616, 1587442168.832, 1587475198.976, 1587475723.264, 
1587497874.432, 1587510064.128, 1587542963.2, 1587559740.416, 
1587577959.424, 1587582022.656, 1587590149.12, 1587596047.36, 
1587634844.672, 1587657257.984, 1587672593.408, 1587711783.936, 
1587761460.224, 1587766572.032, 1587788723.2, 1587831321.6, 1587844297.728, 
1587864876.032, 1587896202.24, 1587915207.68, 1587916387.328, 
1587919401.984, 1587925824.512, 1588001977.344, 1588007613.44, 
1588010234.88, 1588014560.256, 1588028584.96, 1588058076.16, 
1588084814.848, 1588089664.512, 1588105262.08, 1588155986.944, 
1588159525.888, 1588177220.608, 1588243543.04, 1588252455.936, 
1588274344.96, 1588274476.032, 1588275000.32, 1588275524.608, 
1588291384.32, 1588348138.496, 1588393620.48, 1588433204.224, 
1588433728.512, 1588447491.072, 1588448932.864, 1588461122.56, 
1588509357.056, 1588511978.496, 1588518663.168, 1588519711.744, 
1588596651.008, 1588601631.744, 1588608971.776, 1588611462.144, 
1588625093.632, 1588627846.144, 1588696134.656, 1588699411.456, 
1588700591.104, 1588777006.08, 1588777923.584, 1588791686.144, 
1588792865.792, 1588797060.096, 1588855649.28, 1588868363.264, 
1588874392.576, 1588875047.936, 1588882387.968, 1588932588.544, 
1589023552.512, 1589039936.512, 1589078733.824, 1589081355.264, 
1589126444.032, 1589214524.416, 1589230384.128, 1589265511.424, 
1589275866.112, 1589290808.32, 1589292774.4, 1589314139.136, 
1589404054.528, 1589473653.76, 1589476275.2, 1589492397.056, 
1589495280.64, 1589561209.856, 1589603152.896, 1589627270.144, 
1589632381.952, 1589799236.608, 1589806707.712, 1590004626.432, 
1590227448.832, 1590234133.504, 1590398628.864, 1590489068.544, 
1590494180.352, 1590555915.264, 1590707434.496, 1590771266.56, 
1590934451.2, 1591010997.248, 1591383897.088, 1591444452.352, 
1591984468.992, 1592504038.4, 1592553714.688, 1592841679.872, 
1592929629.184, 1592951256.064, 1594128937.984, 1594499347.456, 
1595402171.392, 1595711370.24, 1597937103.872, 1598717768.704, 
1599060521.984, 1599758087.168, 1599815496.704, 1600702198.784, 
1600719631.36, 1601065923.584, 1601119400.96, 1601215476.736, 
1601236710.4, 1601416934.4, 1601499640.832, 1601587328, 1601741206.528, 
1601848423.424, 1601901245.44, 1601913828.352, 1602092872.704, 
1602285417.472, 1602362881.024, 1602504963.072, 1602518987.776, 
1602551231.488, 1602782311.424, 1602783491.072, 1602785457.152, 
1602851124.224, 1602856629.248, 1602964501.504, 1602974594.048, 
1603078140.928), class = c("POSIXct", "POSIXt"), tzone = "UTC")), row.names = c(NA, 
-398L), class = c("tbl_df", "tbl", "data.frame"))

【问题讨论】：

标签： r moving-average rolling-computation summarize

【解决方案1】：

这是data.table 方法。代码说明在代码的cmets中

library( data.table )
#set data to data.table format
setDT( mydata )
#summarise by day
DT <- mydata[, .(total = .N), by = .(date = as.IDate( Admission ), Summary ) ]
#cast to wide
DT <- dcast(DT, date ~ Summary, value.var = "total", fill = 0 )
#create final based on min/max data in DT
final <- data.table( date = seq( min( DT$date ), max( DT$date ), by = 1 ) )
#initialise columns with zero-value
final[, `:=`( Dead = 0, Discharged = 0 ) ]
#update column values based on date
final[ DT, `:=`( Dead = i.Dead, Discharged = i.Discharged ), on = .(date) ][]
#calculate rolling 7 day mortality-rate
final[, mortality.rate := frollsum( Dead, n = 7 ) / ( frollsum( (Dead + Discharged), n = 7 ) ) ]

【讨论】：

嗨！抱歉，这不起作用，因为我不是在寻找每天的平均死亡人数，而是在寻找死亡率（# 死亡患者 / # 死亡和出院患者）。以 4 月 8 日为例，您得到的值为 2.1428，但真实值为 14/39 = 26%
如果我看4月8日，我想知道这天和前六天入院的病人是什么。然后我希望代码意识到该样本中有 x 个死亡患者和 y 个出院患者，并根据 x/x+y 计算死亡率

【解决方案2】：

这里你有一个 tidyverse 方法。我也在使用zoo 包。

计算每天的入院和死亡人数

library(tidyverse)

summaries <- df %>% 
    mutate(date = as.Date(Admission)) %>% 
    group_by(date) %>%
    summarise(total_adm = n(),
              deaths = sum(Summary == "Dead"))

summaries
#> # A tibble: 129 x 3
#>    date       total_adm deaths
#>  * <date>         <int>  <int>
#>  1 2020-01-09         1      0
#>  2 2020-01-27         1      0
#>  3 2020-01-31         2      0
#>  4 2020-02-13         1      0
#>  5 2020-02-21         1      1
#>  6 2020-02-22         1      0
#>  7 2020-02-23         1      0
#>  8 2020-02-24         1      0
#>  9 2020-02-28         1      1
#> 10 2020-03-01         1      1
#> # … with 119 more rows

计算滚动总和和死亡率

对于日期，我首先创建一个序列，然后是 left_join()。在rollsum() 中使用align = "right" 表示您获得前六天。

# sequence fro min to max date
date_seq <- seq.Date(from = min(summaries$date), 
         to = max(summaries$date), 
         by = 1)


tibble(date = date_seq) %>% 
    left_join(summaries) %>% 
    # Calculating rolling sums for admission and deaths, and mortality rate
    mutate(
        adm_7_days = zoo::rollsum(total_adm, k = 7, fill = NA, align = "right", na.rm = TRUE),
        deaths_7_days = zoo::rollsum(deaths, k = 7, fill = NA, align = "right", na.rm = TRUE),
        mortality_rate = deaths_7_days/adm_7_days
    ) %>% 
    print(n = 20)
#> Joining, by = "date"
#> # A tibble: 285 x 6
#>    date       total_adm deaths adm_7_days deaths_7_days mortality_rate
#>    <date>         <int>  <int>      <int>         <int>          <dbl>
#>  1 2020-01-09         1      0         NA            NA             NA
#>  2 2020-01-10        NA     NA         NA            NA             NA
#>  3 2020-01-11        NA     NA         NA            NA             NA
#>  4 2020-01-12        NA     NA         NA            NA             NA
#>  5 2020-01-13        NA     NA         NA            NA             NA
#>  6 2020-01-14        NA     NA         NA            NA             NA
#>  7 2020-01-15        NA     NA          1             0              0
#>  8 2020-01-16        NA     NA          0             0            NaN
#>  9 2020-01-17        NA     NA          0             0            NaN
#> 10 2020-01-18        NA     NA          0             0            NaN
#> 11 2020-01-19        NA     NA          0             0            NaN
#> 12 2020-01-20        NA     NA          0             0            NaN
#> 13 2020-01-21        NA     NA          0             0            NaN
#> 14 2020-01-22        NA     NA          0             0            NaN
#> 15 2020-01-23        NA     NA          0             0            NaN
#> 16 2020-01-24        NA     NA          0             0            NaN
#> 17 2020-01-25        NA     NA          0             0            NaN
#> 18 2020-01-26        NA     NA          0             0            NaN
#> 19 2020-01-27         1      0          1             0              0
#> 20 2020-01-28        NA     NA          1             0              0
#> # … with 265 more rows

【讨论】：

【解决方案3】：

这里的包runner也有帮助

library(dplyr)
library(tidyr)
library(runner)
df %>% 
  group_by(date = as.Date(strptime(Admission, format = "%Y-%m-%d"))) %>%
  pivot_wider(id_cols = date, 
              names_from = Summary, 
              values_from = Admission, 
              values_fn = length, 
              values_fill = 0) %>%
  ungroup() %>%
  complete(date = seq.Date(min(.$date), max(.$date), by = 'day'), 
           fill = list(Discharged = 0, Dead = 0)) %>%
  mutate(Total = Discharged + Dead) %>%
  mutate(mortality_rate = sum_run(Dead, k = 7)/sum_run(Total, k = 7))

# A tibble: 285 x 5
   date       Discharged  Dead Total mortality_rate
   <date>          <dbl> <dbl> <dbl>          <dbl>
 1 2020-01-09          1     0     1              0
 2 2020-01-10          0     0     0              0
 3 2020-01-11          0     0     0              0
 4 2020-01-12          0     0     0              0
 5 2020-01-13          0     0     0              0
 6 2020-01-14          0     0     0              0
 7 2020-01-15          0     0     0              0
 8 2020-01-16          0     0     0            NaN
 9 2020-01-17          0     0     0            NaN
10 2020-01-18          0     0     0            NaN
# ... with 275 more rows

【讨论】：